{ "skills": [ { "id": "skl_agent_schema_drift", "name": "agent-schema-drift", "plugin": "cwc-agent-resources", "description": "The automated cross-language AgentModel drift check CLAUDE.md said didn't exist yet: regex- extracts the enum from all four real mirrors — crates/schema/src/agent.rs (serde, source of truth), scripts/agents/schema.ts (Zod), scripts/agents/schema.py (Pydantic), scripts/agents/schema.java — asserts set-equality, and checks every .claude/agents/*.yaml model: field is a member. Exit nonzero on drift, naming which mirror diverges. Use when editing any of the four mirrors, adding an AgentModel variant, regenerating agent yamls, or on \"check schema drift\", \"are the mirrors in sync\", \"did schema.ts drift again\". Do NOT use for the coworker IDENTITY/BRAND schema (scripts/coworkers/schema.py + plugins/_schema/ validate.py own that), or to fix drift — edit all four mirrors together, then RUSTC_WRAPPER='' cargo run -p agent-gen.", "summary": "The automated cross-language AgentModel drift check CLAUDE.md said didn't exist yet: regex- extracts the enum from all four real mirrors — crates/schema/src/agent.rs (serde, source of truth), scripts/agents/schema.ts (Zod), scripts/agents/schema.py (Pydantic), scripts/agents/schema.java — asserts...", "gated": 0, "source_path": "plugins/cwc-agent-resources/skills/agent-schema-drift/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_agent_schema_drift" }, { "id": "skl_coworker_lifecycle", "name": "coworker-lifecycle", "plugin": "cwc-agent-resources", "description": "The coworker lifecycle for the subagent*.com AI workforce: recruit (when a domain earns a plugin), onboard (the bootstrap checklist), evaluate (rubric rows, not vibes), grow (skills from real work only), dissolve (the founder precedent). Use when creating/retiring a coworker plugin, reviewing coworker performance, changing permission grants, or on \"new coworker\", \"onboard\", \"org design\", \"dissolve plugin\". Org changes are operator-gated.", "summary": "The coworker lifecycle for the subagent*.com AI workforce: recruit (when a domain earns a plugin), onboard (the bootstrap checklist), evaluate (rubric rows, not vibes), grow (skills from real work only), dissolve (the founder precedent).", "gated": 0, "source_path": "plugins/cwc-agent-resources/skills/coworker-lifecycle/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_coworker_lifecycle" }, { "id": "skl_coworkers_console", "name": "coworkers-console", "plugin": "cwc-agent-resources", "description": "The human's localhost entry point to the AI workforce: a standalone-HTML console (apps/coworkers-console/) over the fleet's durable state — coworker IDENTITY/BRAND cards with critters, the e2m mailbox with silence-alerts, and reports (WBR, territory, sweep, toolchain envelopes). Use on \"open the console\", \"show my coworkers\", \"check the mailbox\", or when the operator needs to see coworker output without reading JSONL by hand.", "summary": "The human's localhost entry point to the AI workforce: a standalone-HTML console (apps/coworkers-console/) over the fleet's durable state — coworker IDENTITY/BRAND cards with critters, the e2m mailbox with silence-alerts, and reports (WBR, territory, sweep, toolchain envelopes).", "gated": 0, "source_path": "plugins/cwc-agent-resources/skills/coworkers-console/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_coworkers_console" }, { "id": "skl_analytics_playbook", "name": "analytics-playbook", "plugin": "cwc-analytics-engineering", "description": "analytics-engineering playbook — read real Cloudflare GraphQL Analytics API data for this account's zones and record it as a durable snapshot in subagentcache.com's D1-backed cache_snapshots table, with a manual_dashboard_paste vs graphql_api provenance distinction. Use when the user pastes numbers from dash.cloudflare.com//analytics, asks to check cache hit rate, bandwidth, request counts, or wants a live GraphQL Analytics query run. Trigger on \"cache analytics\", \"cloudflare analytics\", \"cache hit rate\", \"cached bandwidth\", \"check the dashboard\", \"graphql analytics query\", \"wire up live analytics\", \"httpRequests1dGroups\", or a pasted Cloudflare Analytics dashboard screenshot/ text. Do NOT use this to invent or auto-expand a GraphQL query the user didn't ask for — the underlying Cloudflare GraphQL tool is explicitly restricted to exact user-provided queries only; see the \"Hard constraint\" section before writing any query.", "summary": "analytics-engineering playbook — read real Cloudflare GraphQL Analytics API data for this account's zones and record it as a durable snapshot in subagentcache.com's D1-backed cache_snapshots table, with a manual_dashboard_paste vs graphql_api provenance distinction.", "gated": 1, "source_path": "plugins/cwc-analytics-engineering/skills/analytics-playbook/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_analytics_playbook" }, { "id": "skl_credential_proxy", "name": "credential-proxy", "plugin": "cwc-analytics-engineering", "description": "Run a local credential-injecting proxy, per the pattern documented at code.claude.com/docs/en/agent-sdk/secure-deployment#credential-management, so Claude can pull data from a credentialed API (starting with Cloudflare's GraphQL Analytics API) without ever touching the plaintext token itself -- the proxy reads Keychain and injects the header; Claude sends and receives credential-free requests through it. Use when an MCP tool that would normally provide this access is unavailable/down and the operator wants Claude to complete the pull anyway via a locally-run proxy instead. Trigger on \"set up a proxy\", \"credential proxy\", \"inject credentials safely\", \"the MCP server is down, use a proxy instead\". Do NOT use this to bypass the fivetran-keychain-secrets Hard constraint -- see this file's own Hard constraint section for the one boundary that still applies.", "summary": "Run a local credential-injecting proxy, per the pattern documented at code.claude.com/docs/en/agent-sdk/secure-deployment#credential-management, so Claude can pull data from a credentialed API (starting with Cloudflare's GraphQL Analytics API) without ever touching the plaintext token itself -- t...", "gated": 0, "source_path": "plugins/cwc-analytics-engineering/skills/credential-proxy/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_credential_proxy" }, { "id": "skl_data_vision_runbook", "name": "data-vision-runbook", "plugin": "cwc-analytics-engineering", "description": "The durable \"why + where we are\" document for .coworkers-analytics's multi-connector Fivetran data-infra initiative — why this repo is building its own connector-backed data layer, what \"done\" looks like for the rollout, and the exact current state of every connector (registry row counts, which are fivetran-debug-verified, which are real vs. best-guess schemas, which are wired to live data vs. zero-row skeleton). Use when the user asks \"what's the data vision\", \"where are we on Fivetran connectors\", \"why are we building this\", \"what's done vs. not done on the connector rollout\", \"which connectors are verified\", \"recap the data-infra state\", or wants to know current registry numbers before deciding what to build next. Trigger on \"data vision\", \"connector rollout state\", \"fivetran status\", \"connector registry\", \"coworkers-analytics data layer\". Do NOT use this for breaking the rollout into atomic tasks/tickets — that's `plugins/cwc-product-management/skills/fivetran-rollout-decomposition`. Do NOT use this to execute code changes, write connector.py files, edit schema.ts, or run migrations — that's `plugins/cwc-engineering/skills/engineering-playbook`, which hands off to cwc-deploy with the exact phrase \"ready for cwc-deploy\" once code is written. This skill only states and maintains the vision and the ground-truth snapshot; it does not plan tasks and does not touch a keyboard.", "summary": "The durable \"why + where we are\" document for .coworkers-analytics's multi-connector Fivetran data-infra initiative — why this repo is building its own connector-backed data layer, what \"done\" looks like for the rollout, and the exact current state of every connector (registry row counts, which a...", "gated": 0, "source_path": "plugins/cwc-analytics-engineering/skills/data-vision-runbook/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_data_vision_runbook" }, { "id": "skl_deploy_wave_analytics", "name": "deploy-wave-analytics", "plugin": "cwc-analytics-engineering", "description": "Per-wave timing analytics over deploy-runner step logs: parses the `STEP=n NAME=x EXIT=rc DURATION_S=d` / terminal `RUN_EXIT=rc` line format owned by plugins/cwc-deploy/skills/deploy-runner/deploy_runner/log.py (byte-identical regexes, enforced by a drift test that re-parses log.py at test time) into a typed `WaveReport` — step count, total/mean/max duration, slowest step name — via `uv run wave-analytics `. Mirrors log.py's hard rule exactly: success comes ONLY from an explicit `RUN_EXIT=0`; an unterminated log is RUNNING, never SUCCESS. This skill READS logs, never writes them. Use when analyzing how long a deploy wave took, which step was slowest, comparing waves, or checking whether a wave log actually terminated. Trigger on \"wave timing\", \"slowest deploy step\", \"analyze the deploy log\", \"how long did wave N take\". Do NOT use to RUN deploys or generate step logs (cwc-deploy's deploy-runner owns that end) or for Cloudflare cache/CDN analytics (sibling analytics-playbook owns subagentcache.com).", "summary": "Per-wave timing analytics over deploy-runner step logs: parses the `STEP=n NAME=x EXIT=rc DURATION_S=d` / terminal `RUN_EXIT=rc` line format owned by plugins/cwc-deploy/skills/deploy-runner/deploy_runner/log.py (byte-identical regexes, enforced by a drift test that re-parses log.py at test time)...", "gated": 0, "source_path": "plugins/cwc-analytics-engineering/skills/deploy-wave-analytics/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_deploy_wave_analytics" }, { "id": "skl_fivetran_key_rotation", "name": "fivetran-key-rotation", "plugin": "cwc-analytics-engineering", "description": "Rotate a Fivetran or connector-source API key stored via the fivetran-keychain-secrets skill, for scheduled rotation or an emergency (suspected leak/exposure). Use when the user asks to rotate, refresh, or regenerate a Fivetran secret, or reports a key may have been exposed. Trigger on \"rotate the key\", \"regenerate the secret\", \"the fivetran key leaked\", \"emergency rotation\". This skill prints a checklist; it does not perform the rotation itself -- see the Hard constraint section for why.", "summary": "Rotate a Fivetran or connector-source API key stored via the fivetran-keychain-secrets skill, for scheduled rotation or an emergency (suspected leak/exposure).", "gated": 0, "source_path": "plugins/cwc-analytics-engineering/skills/fivetran-key-rotation/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_fivetran_key_rotation" }, { "id": "skl_fivetran_keychain_secrets", "name": "fivetran-keychain-secrets", "plugin": "cwc-analytics-engineering", "description": "Store Fivetran/connector-source API keys in the macOS login Keychain instead of plaintext files or Claude's own context, following the same \"encrypted at rest, never touched by the untrusted process\" pattern .coworkers-analytics uses for its own credentials (encryption.key + .oauth-tokens/*.enc). Use when the user wants to set up, store, or rotate a Fivetran API key or a connector source credential (Cloudflare API token, GitHub PAT, etc). Trigger on \"store the fivetran key\", \"keychain\", \"set up fivetran secrets\", \"rotate the api key\". Do NOT use this skill to type, echo, grep for, or otherwise handle the plaintext secret value yourself in any tool call -- see the Hard constraint section.", "summary": "Store Fivetran/connector-source API keys in the macOS login Keychain instead of plaintext files or Claude's own context, following the same \"encrypted at rest, never touched by the untrusted process\" pattern .coworkers-analytics uses for its own credentials (encryption.key + .oauth-tokens/*.enc).", "gated": 0, "source_path": "plugins/cwc-analytics-engineering/skills/fivetran-keychain-secrets/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_fivetran_keychain_secrets" }, { "id": "skl_network_analytics", "name": "network-analytics", "plugin": "cwc-analytics-engineering", "description": "Visualize and track the network metrics that cwc-hardware-engineering's network-capture skill records — reads the SAME network-event JSONL / SQLite fact-dim star schema (the shared contract) and produces polars rollups + a deterministic inline SVG dashboard. `uv run netviz rollup --jsonl

` (latency percentiles by resource_type, bytes by host, totals — as JSON), `uv run netviz svg --jsonl

` (byte-identical SVG in the house server-side-SVG style). Use when the user asks to visualize / chart / analyze / track captured network data, see p50/p95 latency, top talkers by bytes, or a request-type breakdown. Trigger on \"visualize network\", \"network dashboard\", \"latency percentiles\", \"bytes by host\", \"analyze captured requests\", \"network chart\". Do NOT use to CAPTURE data (cwc-hardware-engineering's network-capture writes what this reads) or for Cloudflare cache analytics (analytics-playbook owns subagentcache.com).", "summary": "Visualize and track the network metrics that cwc-hardware-engineering's network-capture skill records — reads the SAME network-event JSONL / SQLite fact-dim star schema (the shared contract) and produces polars rollups + a deterministic inline SVG dashboard.", "gated": 0, "source_path": "plugins/cwc-analytics-engineering/skills/network-analytics/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_network_analytics" }, { "id": "skl_fleet_inbox_digest", "name": "fleet-inbox-digest", "plugin": "cwc-customer-support", "description": "Scheduled daily support digest for the subagent*.com fleet — one pass over the fleet-wide health fan-out (subagentapi.com/health) and the open durable task rows (subagenttasks.com/api/tasks), reproducing any NEW failure with curl before recording it, ending in exactly one digest file the operator reads (reports/support/digest-YYYYMMDD.md). Use when asked for a \"support digest\", \"daily triage\", \"inbox pass\", \"fleet health digest\", \"morning support pass\", or when a scheduled routine fires this coworker's daily check. This is the SCHEDULED complement to support-playbook (reactive, per-report triage) — do NOT use this skill to answer a single user's live issue report; use support-playbook for that. Do NOT use for fixing bugs (engineering) or deploying (cwc-deploy).", "summary": "Scheduled daily support digest for the subagent*.com fleet — one pass over the fleet-wide health fan-out (subagentapi.com/health) and the open durable task rows (subagenttasks.com/api/tasks), reproducing any NEW failure with curl before recording it, ending in exactly one digest file the operator...", "gated": 0, "source_path": "plugins/cwc-customer-support/skills/fleet-inbox-digest/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_fleet_inbox_digest" }, { "id": "skl_llms_accuracy_check", "name": "llms-accuracy-check", "plugin": "cwc-customer-support", "description": "Verify that a site's documented help surface is TRUE: parse its /llms.txt (the family convention — '# domain — title', '> description', '## API' route bullets, '## Family'), extract every 'GET /api/...' route it promises, request each one live, and classify the site ACCURATE / OVERPROMISING (a documented route 404s) / UNDERDOCUMENTED (an allow-listed conventional route answers but isn't documented — hint only, never a crawl). Use when a user reports \"the docs say this endpoint exists but it 404s\", when auditing a site's llms.txt after a deploy, when support-playbook triage needs to know whether the documentation or the site is wrong, or on a periodic docs-accuracy sweep. Do NOT use for fleet-wide llms.txt COVERAGE metrics (scripts/flywheel-metrics.sh and cwc-marketing own that), for fixing a wrong llms.txt (engineering + cwc-deploy), or for probing undocumented routes beyond the declared allow-list — this skill never crawls.", "summary": "Verify that a site's documented help surface is TRUE: parse its /llms.txt (the family convention — '# domain — title', '> description', '## API' route bullets, '## Family'), extract every 'GET /api/...' route it promises, request each one live, and classify the site ACCURATE / OVERPROMISING (a do...", "gated": 0, "source_path": "plugins/cwc-customer-support/skills/llms-accuracy-check/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_llms_accuracy_check" }, { "id": "skl_support_playbook", "name": "support-playbook", "plugin": "cwc-customer-support", "description": "Customer-support playbook for the subagent*.com fleet — triage issue reports against live health checks (subagentapi.com/health, the affected site's own API root), reproduce with curl before answering anything, file real bugs as durable task rows on subagenttasks.com with the failing curl evidence, and apply known-issue awareness (a brief 404/1104 right after a deploy is stale edge cache, not a code bug). Use when the user reports \"site X is down\", \"the API returns an error\", \"this endpoint is broken\", \"a user hit a 404\", \"is the fleet healthy\", \"triage this\", \"reproduce this bug\", or \"file this issue\". Do NOT use for fixing the bug itself (that is engineering work through the normal repo flow) or for deploying (cwc-deploy). Do NOT confuse subagentsupport.com with a ticket system — it is a read-only catalog documenting Anthropic's public customer-support plugin.", "summary": "Customer-support playbook for the subagent*.com fleet — triage issue reports against live health checks (subagentapi.com/health, the affected site's own API root), reproduce with curl before answering anything, file real bugs as durable task rows on subagenttasks.com with the failing curl evidenc...", "gated": 0, "source_path": "plugins/cwc-customer-support/skills/support-playbook/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_support_playbook" }, { "id": "skl_data_playbook", "name": "data-playbook", "plugin": "cwc-data", "description": "Data coworker playbook — the single source of truth for safe and effective querying of subagentjobs-dwh (Cloudflare D1, Kimball star schema). Use when the user asks a question about job posting data, the crawler pipeline, board freshness, skills/departments/offices, or wants SQL written/validated against this warehouse. Trigger on \"how many jobs\", \"query the warehouse\", \"write a query for\", \"is this SQL right\", \"which boards are stale\", \"what's the grain of X table\", or any follow-up analytics question after /data. Do NOT use for editing workers/cron or workers/etl source code (that's a coding task, not an analytics question) or for questions about data outside subagentjobs-dwh.", "summary": "Data coworker playbook — the single source of truth for safe and effective querying of subagentjobs-dwh (Cloudflare D1, Kimball star schema).", "gated": 0, "source_path": "plugins/cwc-data/skills/data-playbook/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_data_playbook" }, { "id": "skl_weekly_business_review", "name": "weekly-business-review", "plugin": "cwc-data", "description": "Run the weekly business review (WBR) and monthly business review (MBR) for the subagent*.com business: input metrics before output metrics, exceptions before averages, and — the point — ranked, forward-looking recommendations with predicted metric movement and a named gate owner. Use on a weekly cadence, when the operator says \"WBR\", \"business review\", \"what do the numbers say\", or when any strategy/budget artifact needs a data check. Tracking without a recommendation is a dashboard, not a review.", "summary": "Run the weekly business review (WBR) and monthly business review (MBR) for the subagent*.com business: input metrics before output metrics, exceptions before averages, and — the point — ranked, forward-looking recommendations with predicted metric movement and a named gate owner.", "gated": 0, "source_path": "plugins/cwc-data/skills/weekly-business-review/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_weekly_business_review" }, { "id": "skl_zones_query_pack", "name": "zones-query-pack", "plugin": "cwc-data", "description": "Read-only jq query pack over the family's canonical DNS-zone registry API (https://subagentsubdomains.com/api/dns-zones — the machine-readable directory every site's robots.txt points agents at). Six named recipes: counts by status, subagent*/agent*/other name-family split, NULL-zone_id rows, launch batches by created_at date, never_use/tombstone rows, newest N. Wrapper runs each recipe against the live endpoint (--live) or the committed 2026-07-09 fixture (default). Use when the user asks how many zones/domains the fleet has, which are routed vs workers.dev-only, which zones lack a zone_id, when a launch wave shipped, which domains are tombstoned/never-use, or what launched most recently. Trigger on \"dns zones\", \"how many domains\", \"zone registry\", \"launch batch\", \"tombstone\", \"never use domain\". Do NOT use for subagentjobs-dwh warehouse questions (data-playbook owns that schema), for editing the registry (this pack is read-only — writes go through workers/subagentsubdomains' gated API), or for Cloudflare zone-ID lookups by API token (that's scripts/cf-zone-id.sh).", "summary": "Read-only jq query pack over the family's canonical DNS-zone registry API (https://subagentsubdomains.com/api/dns-zones — the machine-readable directory every site's robots.txt points agents at).", "gated": 0, "source_path": "plugins/cwc-data/skills/zones-query-pack/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_zones_query_pack" }, { "id": "skl_atlas_telemetry_publisher", "name": "atlas-telemetry-publisher", "plugin": "cwc-data-engineering", "description": "Publish the Coworkers board's session telemetry into the REAL Netflix Atlas standalone server on :7101 (tsk_atlas_publish_telemetry, W2 of the Atlas fleet plan). Reads the frozen SessionEvent JSONL wire shape (trace_id/span_id/coworker/sprint_id/kind/ts) from `~/Library/Application Support/Coworkers/telemetry.jsonl`, aggregates counts per UTC step bucket per (coworker, sprintId, kind), and plain-POSTs `/api/v1/publish` payloads (`{\"tags\":{\"name\":\"coworker.session.events\"},\"metrics\":[...]}`) — no spectator-py. Two modes: `uv run atlas-publish replay` (full-history backfill; chronological per-series order is MANDATORY — Atlas silently drops out-of-order older points with HTTP 200, live-verified) and `uv run atlas-publish follow --interval 60` (foreground loop, closed buckets only; no daemon — launchd is a gated doc-only note). Source-list design: W5's cf.requests source plugs in additively. KNOWN SERVER LIMITS (live-verified 2026-07-05, see README table): graph steps >1h return empty legend under the current block-size-60 conf (step=1d presets need W1 conf change or W3 query fallback); consolidation averages so append `:cf-sum` to read raw counts; query index lags publishes by 10-60s; backfill accepted to >=14d. Use for publishing/backfilling fleet telemetry into Atlas, debugging why a published series doesn't graph, or adding a second metric source. Trigger on \"atlas publish\", \"replay telemetry\", \"backfill atlas\", \"follow mode\", \"publish payload\", \"cf-sum\", \"coworker.session.events\". Do NOT use for running the Atlas server itself (scripts/atlas/run-atlas.sh, cwc-engineering), the graph-API prototype (cwc-engineering's atlas-graph-prototype), preset/manifest changes (W3, schemas/atlas-graph-presets.json), or the star-schema ERD (session-telemetry-erd).", "summary": "Publish the Coworkers board's session telemetry into the REAL Netflix Atlas standalone server on :7101 (tsk_atlas_publish_telemetry, W2 of the Atlas fleet plan).", "gated": 0, "source_path": "plugins/cwc-data-engineering/skills/atlas-telemetry-publisher/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_atlas_telemetry_publisher" }, { "id": "skl_corpus_cleaning_pipeline", "name": "corpus-cleaning-pipeline", "plugin": "cwc-data-engineering", "description": "Typed models (Pydantic + FilterReason enum) and verified, profiled performance/library-quality reference notes for an offline LLM-pretraining-corpus cleaning pipeline — schema validation, vectorized quality filters, exact-hash dedup, datasketch-backed MinHash+LSH near-dedup, real tiktoken token counting, a composite quality-ranking score, sharded parquet output. Grounded in the real pipeline.py at organizations/workspaces/projects/2026-07-06-xai-interview-prep/ take-home/submission/ (an xAI Data Engineer take-home submission), not a hypothetical framework — this skill's SKILL.md and models.py cite that file's actual stage names, actual measured thresholds, and its actual library choices. `FilterReason`'s enum members are grepped 1:1 from pipeline.py's real filter-name strings (tests/test_models.py enforces this as a live drift check, not a static claim) — if pipeline.py's filters change, this skill's tests fail until the enum is updated to match. Deliberately does NOT add pydantic as a runtime dependency to pipeline.py itself; this skill is the typed documentation/verification layer alongside it, scope-honestly framed as a DELIBERATE broadening of cwc-data-engineering's charter beyond its original Coworkers session-telemetry domain (user-approved explicitly, not an organic extension — see README.md's \"Scope\" section). Use when working on corpus/document-cleaning batch pipelines, MinHash/LSH near-dedup, tiktoken-based token accounting, or the xAI take-home submission specifically. Do NOT use for Coworkers board/session telemetry (sibling `session-telemetry-erd` skill's territory) or Cloudflare cache/CDN snapshots (cwc-analytics-engineering).", "summary": "Typed models (Pydantic + FilterReason enum) and verified, profiled performance/library-quality reference notes for an offline LLM-pretraining-corpus cleaning pipeline — schema validation, vectorized quality filters, exact-hash dedup, datasketch-backed MinHash+LSH near-dedup, real tiktoken token c...", "gated": 0, "source_path": "plugins/cwc-data-engineering/skills/corpus-cleaning-pipeline/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_corpus_cleaning_pipeline" }, { "id": "skl_ledger_quality", "name": "ledger-quality", "plugin": "cwc-data-engineering", "description": "Data-quality checks over the REAL fleet-deploy ledger (`fact_fleet_deploy` in local Postgres, migration `crates/durable-store/migrations/postgres/ 008_fleet_deploy_ledger.sql`) — typed rows (Pydantic `LedgerRow` + a `DeployStatus` enum whose 6 members are the migration's own `CREATE TYPE deploy_status` values, enforced by a live drift test that re-parses the SQL at test time), pure-function invariant checks (all-terminal after a completed batch, unique (domain, wave), verified rows carry verified_at, skipped rows carry notes), and a thin psql-backed CLI (`uv run ledger-quality check`) for live runs. Use when auditing a deploy wave's ledger state after a batch, validating ledger rows from any non-Postgres source (fixtures, CSV export, a merge), or checking the ledger before a closeout report. Do NOT use to WRITE ledger rows (plugins/cwc-deploy/skills/fleet-expansion/ scripts/ledger.sh is the only writer) or for deploy step-log timing analytics (cwc-analytics-engineering's deploy-wave-analytics owns the wave logs).", "summary": "Data-quality checks over the REAL fleet-deploy ledger (`fact_fleet_deploy` in local Postgres, migration `crates/durable-store/migrations/postgres/ 008_fleet_deploy_ledger.sql`) — typed rows (Pydantic `LedgerRow` + a `DeployStatus` enum whose 6 members are the migration's own `CREATE TYPE deploy_s...", "gated": 0, "source_path": "plugins/cwc-data-engineering/skills/ledger-quality/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_ledger_quality" }, { "id": "skl_parquet_shape_profiler", "name": "parquet-shape-profiler", "plugin": "cwc-data-engineering", "description": "Compute a structured, typed data-shape profile (DatasetShapeProfile: row/column counts, file size, per-column dtype + kind + null rate, numeric percentiles [min/p25/median/p75/max/mean] or categorical distinct-count + top-values) for any parquet dataset, as validated Pydantic JSON — `uv run shape-profiler ` (add `--out profile.json` to write to a file). Every number in the output is a REAL computed statistic, never a placeholder — the same \"capture is source-shaped, never invented\" discipline the sibling corpus-cleaning-pipeline skill applies to FilterReason. This skill owns COMPUTING the numbers; the sibling plugin cwc-design's `data-shape-svg` skill owns rendering the resulting JSON as SVG markup — a deliberate cross-plugin split (profiler computes, design draws) first proven end-to-end on the xAI take-home's real 100MB sample and rendered live on agenttables.com. Use when the user wants to profile a parquet file's shape/structure, needs real column statistics before designing filters or visualizations, or asks \"what does this data look like\". Do NOT use to render the visualization itself (data-shape-svg's territory), to clean/filter the data (corpus-cleaning-pipeline), or for Coworkers board/session telemetry (session-telemetry-erd).", "summary": "Compute a structured, typed data-shape profile (DatasetShapeProfile: row/column counts, file size, per-column dtype + kind + null rate, numeric percentiles [min/p25/median/p75/max/mean] or categorical distinct-count + top-values) for any parquet dataset, as validated Pydantic JSON — `uv run shape...", "gated": 0, "source_path": "plugins/cwc-data-engineering/skills/parquet-shape-profiler/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_parquet_shape_profiler" }, { "id": "skl_session_telemetry_erd", "name": "session-telemetry-erd", "plugin": "cwc-data-engineering", "description": "The session-telemetry ERD and durable data spine under the Coworkers sprint board — TWO layers, fivetran-style. (1) The CONNECTOR layer (`connector_schema.sql` + connector.py): source-shaped tables named after the REAL fivetran claude-compliance/claude-platform connector entities (organization, chat, chat_message, chat_message_usage w/ exact message_usage_report_result token columns, project, model, skill, claude_code_usage_report) plus the Cowork SDK's own on-disk shapes (cowork_vm_session from bridge-state.json's environmentId/remoteSessionId/processedMessageUuids), scoped to what a claude.ai 20x subscription can ACTUALLY capture — everything else lives in the UNAVAILABLE ledger (rendered as the ERD's dark legend callout) with the API each table would need. `uv run telemetry-erd connector-sync --db [--plugins …] [--email …]` syncs the real Mac sources; `connector-erd --out ` renders the fivetran-grammar ERD; `connector-schema` prints the DDL. (2) The MART layer (`schema.sql`): Kimball fact_/dim_ star schema (fact_task/fact_session_event/dim_coworker/dim_sprint/dim_status) derived downstream for the sprint board, with the cross-language PARITY guard (Swift CoworkersModel ↔ task-session JSON schema ↔ SQL). `load --board ` / `schema` as before. NEVER invent fact_/dim_ names at the capture layer — capture is source-shaped; marts are derived. Use for session/board/telemetry modeling, the claude.ai connector equivalent, ERDs, cross-language parity. Trigger on \"connector\", \"fivetran\", \"ERD\", \"star schema\", \"fact table\", \"Kimball\", \"session telemetry\", \"cowork vm data\", \"usage report\". Do NOT use for WBR analysis (cwc-data), Cloudflare snapshots (cwc-analytics-engineering), or the Swift UI (cwc-swift-engineering).", "summary": "The session-telemetry ERD and durable data spine under the Coworkers sprint board — TWO layers, fivetran-style.", "gated": 0, "source_path": "plugins/cwc-data-engineering/skills/session-telemetry-erd/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_session_telemetry_erd" }, { "id": "skl_wbr_warehouse", "name": "wbr-warehouse", "plugin": "cwc-data-engineering", "description": "Materialize and refresh the weekly-business-review (WBR) semantic star schema and its deterministic collectors. Use when asked to collect wbr data, load the warehouse, materialize the warehouse, run monday data readiness, or wire a new WBR data source.", "summary": "Materialize and refresh the weekly-business-review (WBR) semantic star schema and its deterministic collectors.", "gated": 0, "source_path": "plugins/cwc-data-engineering/skills/wbr-warehouse/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_wbr_warehouse" }, { "id": "skl_deploy_playbook", "name": "deploy-playbook", "plugin": "cwc-deploy", "description": "Deploy coworker playbook — the Desktop-Commander-based workflow for scaffolding, installing, type-checking, deploying, and verifying a Cloudflare Worker from a Claude Cowork session, without poisoning the real Mac's node_modules or guessing at gotchas that already cost a prior session real time. Use when the user asks to deploy a Worker, run `wrangler deploy`, set a secret, create or migrate a D1 database, scaffold a new `workers/` project, or verify a domain went live. Trigger on \"deploy this\", \"ship it\", \"wrangler deploy\", \"set the secret\", \"create a D1 database\", \"is it live\", \"scaffold a new worker\", or any Desktop-Commander `interact_with_process` session touching npm/tsc/wrangler. Do NOT use this to run npm/tsc/wrangler directly from a cloud sandbox shell — that is the one thing this playbook exists to prevent. Do NOT use for application code changes themselves (routing, HTML, SQL logic) — this is the ship step, not the build step.", "summary": "Deploy coworker playbook — the Desktop-Commander-based workflow for scaffolding, installing, type-checking, deploying, and verifying a Cloudflare Worker from a Claude Cowork session, without poisoning the real Mac's node_modules or guessing at gotchas that already cost a prior session real time.", "gated": 1, "source_path": "plugins/cwc-deploy/skills/deploy-playbook/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_deploy_playbook" }, { "id": "skl_deploy_runner", "name": "deploy-runner", "plugin": "cwc-deploy", "description": "Executable tool that turns deploy-playbook's Desktop-Commander session-corruption guidance into tested code, not just prose. Use it whenever a deploy sequence contains a step of unknown or uncertain duration (a first-time npm install, a cold wrangler deploy, anything that might run long) — it renders a step-logging bash script, launches it backgrounded via one short call, and reads success/failure from an explicit log marker rather than session liveness. Trigger on \"deploy this and it might take a while\", \"background the deploy\", \"the session keeps dying/wedging mid-deploy\", or any deploy-playbook step where you're unsure how long a command will run. Do NOT use this for a quick, known-fast command (a `curl` verify, a one-line D1 SELECT) — deploy-playbook's plain Desktop-Commander pattern is simpler and fine there. This tool does not execute wrangler/npm/tsc itself and does not bypass any gate in gates.toml — it only generates the artifact (the script) and interprets the artifact it produces (the log); the actual commands still run through Desktop Commander or the user's own terminal, same as deploy-playbook requires.", "summary": "Executable tool that turns deploy-playbook's Desktop-Commander session-corruption guidance into tested code, not just prose.", "gated": 0, "source_path": "plugins/cwc-deploy/skills/deploy-runner/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_deploy_runner" }, { "id": "skl_edge_cache_doctor", "name": "edge-cache-doctor", "plugin": "cwc-deploy", "description": "Post-deploy diagnosis of the REAL stale-edge-cache-after-deploy pattern this fleet keeps hitting (subagentpython.com 2026-07-01: stale sibling content + error 1104 right after zone activation, self-resolved in seconds; subagentsystemdesigns.com 2026-07-09: first verify saw the old 725-byte parked placeholder, the retry got the 13237-byte worker body). Probes a domain N times with cache-busting queries + no-cache headers, records status/size/cf-cache-status/marker per attempt, and classifies: HEALTHY / STALE-EDGE (some attempt shows the new content — wait and re-probe before diagnosing anything) / ROUTE-MISS (consistent old content, no worker evidence — check wrangler routes, waiting won't help) / DOWN, each with its next action. Use when a just-deployed site \"still shows the old page\", a verify step failed right after wrangler deploy, curl returns a 725/1008- byte placeholder, or before filing any \"deploy didn't take\" diagnosis. Do NOT use as the pass/fail deploy gate — that is ../fleet-expansion/scripts/verify-site.sh's job (this is the diagnosis half, that is the gate half) — and do NOT use it to fix routes (route work is a deploy mutation, governed by this plugin's existing gates).", "summary": "Post-deploy diagnosis of the REAL stale-edge-cache-after-deploy pattern this fleet keeps hitting (subagentpython.com 2026-07-01: stale sibling content + error 1104 right after zone activation, self-resolved in seconds; subagentsystemdesigns.com 2026-07-09: first verify saw the old 725-byte parked...", "gated": 0, "source_path": "plugins/cwc-deploy/skills/edge-cache-doctor/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_edge_cache_doctor" }, { "id": "skl_fleet_expansion", "name": "fleet-expansion", "plugin": "cwc-deploy", "description": "Batch-scaffold, deploy, and register MANY new Cloudflare Worker sites at once across the subagent*.com / agent*.com family, coordinated through a Postgres deploy ledger instead of conversation context. Use when launching workers for multiple owned-but-unserved domains in one pass — \"deploy workers for the domains I own\", \"launch the rest of the fleet\", \"fill the unserved zones\". Composes (never duplicates) the single-site skills: cwc-engineering:worker-build builds ONE worker, cwc-deploy:deploy-playbook/deploy-runner ship ONE worker — this skill owns only what batching adds: the typed fleet manifest, up-front accent allocation, the wave protocol with its serialization rules, the fact_fleet_deploy ledger, and batch verification. Do NOT use for a single new site (deploy-playbook alone is simpler) and do NOT let any wave edit workers/_shared/kit/ (a kit edit mid-batch is a 66-lock fleet relock event).", "summary": "Batch-scaffold, deploy, and register MANY new Cloudflare Worker sites at once across the subagent*.com / agent*.com family, coordinated through a Postgres deploy ledger instead of conversation context.", "gated": 0, "source_path": "plugins/cwc-deploy/skills/fleet-expansion/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_fleet_expansion" }, { "id": "skl_accent_allocator", "name": "accent-allocator", "plugin": "cwc-design", "description": "Automated accent allocation for new subagent*.com sites: a census parser over the real SITES[] registry (workers/subagentdomains/src/index.ts, 94 entries / 83 unique hexes as of 2026-07-09), the family's real hex/255 P3 derivation, and a deterministic largest-hue-gap proposer that emits N visually-spread AccentSpec candidates avoiding every taken hex — validated against the sibling fleet-design-system skill's AccentSpecSchema by reference, not a mirror. Use when the user wants to \"pick an accent\" for a new site, \"allocate accents\" for a fleet-expansion batch, compute a hex's display-p3 value, or census which accents are already taken. Do NOT use for validating an already-chosen batch (that's the collision GATE — plugins/cwc-deploy/skills/fleet-expansion/scripts/check-accents.sh — this skill proposes, the gate validates), for editing the six legacy hand-tuned shared kit tokens in workers/_shared/kit/css.ts (they deliberately do NOT follow the hex/255 formula), or for the token drift check (sibling fleet-design-system skill).", "summary": "Automated accent allocation for new subagent*.com sites: a census parser over the real SITES[] registry (workers/subagentdomains/src/index.ts, 94 entries / 83 unique hexes as of 2026-07-09), the family's real hex/255 P3 derivation, and a deterministic largest-hue-gap proposer that emits N visuall...", "gated": 0, "source_path": "plugins/cwc-design/skills/accent-allocator/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_accent_allocator" }, { "id": "skl_agent_dx_cli_scale", "name": "agent-dx-cli-scale", "plugin": "cwc-design", "description": "A scoring scale for evaluating how well a CLI is designed for AI agents, based on the \"Rewrite Your CLI for AI Agents\" principles.", "summary": "A scoring scale for evaluating how well a CLI is designed for AI agents, based on the \"Rewrite Your CLI for AI Agents\" principles.", "gated": 0, "source_path": "plugins/cwc-design/skills/agent-dx-cli-scale/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_agent_dx_cli_scale" }, { "id": "skl_animation", "name": "animation", "plugin": "cwc-design", "description": "Produce the \"Animation\" project shape (one of claude.ai/design's 5 canonical shapes — Prototype/Slides/Document/Wireframe/Animation) using Motion (motion.dev, npm package \"motion\", CDN-loadable with zero build step — https://cdn.jsdelivr.net/npm/motion@12.42.2/+esm), not hand-rolled CSS keyframes. Covers Motion's real capabilities: independent transforms (x/y/rotate/scale via animate()), scroll-linked animation (scroll()+ScrollTimeline), native gestures (whileHover/whileTap/drag), layout animation, spring physics (type:\"spring\"), exit animation (keep-in-DOM-until-exit pattern), timeline sequences (variants+stagger), and motion values (live-derived state). Two output modes: (1) frontend Workers animation — real Motion-driven CSS/JS running live in a Cloudflare Worker-served page; (2) rendered-video-loop content — a sequence of still images (e.g. higgsfield-generated frames) played back as a looping filmstrip, for cases where the content itself is pre-rendered art rather than DOM elements. Visual language for both modes is this family's own real entity-graph / node-link \"galaxy\" style (glowing cyan/green nodes, thin connecting edges on black — see plugins/cwc-design/generated/ and the real deck.ts hero-band/close-band illustrations), not invented per-project. Use when the user wants a real animation/motion artifact, references Motion/motion.dev, wants a claude.ai/design-style Animation output, or wants the Global Workspace storyboard turned into actual running motion. Do NOT use for static Prototype/Wireframe mockups (sibling skills) or the fleet's own baseline `@view-transition`/reduced-motion CSS rules (fleet-design-system's \"Do's and Don'ts\") — though every animation this skill produces MUST still respect `prefers-reduced-motion`.", "summary": "Produce the \"Animation\" project shape (one of claude.ai/design's 5 canonical shapes — Prototype/Slides/Document/Wireframe/Animation) using Motion (motion.dev, npm package \"motion\", CDN-loadable with zero build step — https://cdn.jsdelivr.net/npm/motion@12.42.2/+esm), not hand-rolled CSS keyframes.", "gated": 0, "source_path": "plugins/cwc-design/skills/animation/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_animation" }, { "id": "skl_architecture_wireframes", "name": "architecture-wireframes", "plugin": "cwc-design", "description": "CRUD operations on the family's architecture diagram / wireframe set across TWO render targets — Figma/FigJam (via the official Figma MCP server) and, since 2026-07-05, the native Swift `FleetGraphView` in src/coworkers (the coworker-PLUGIN dependency layer: which cwc-* plugin depends on which, generated from plugins/cwc-*/DEPENDENCIES.json into schemas/fleet-graph.json — see schemas/FLEET-GRAPH.md). Use when the user wants an architecture diagram, system visualization, wireframe, or FigJam board created, read back, updated, or retired — \"diagram the fleet\", \"visualize the plugin architecture\", \"update the architecture board\", \"screenshot the wireframe\", \"add the new primitive to the diagram\" — or when a Swift change touches FleetGraphView and needs a design-system QA sign-off before it ships (8-token conformance, Liquid Glass conventions, GlassEffectContainer grouping, WCAG contrast) — or pastes a figma.com URL for one of this family's boards. Pairs with cwc-engineering (system facts: workers, crates, schema; also owns fleet-graph-build.py) and cwc-data (warehouse facts) as the sources of what to draw. Do NOT use for the 8-token Hardware Buddy character-pack loop (that's the sibling design-playbook skill), and do NOT invent architecture — every box and edge must be a real worker, plugin, coworker, or API that exists in the repo.", "summary": "CRUD operations on the family's architecture diagram / wireframe set across TWO render targets — Figma/FigJam (via the official Figma MCP server) and, since 2026-07-05, the native Swift `FleetGraphView` in src/coworkers (the coworker-PLUGIN dependency layer: which cwc-* plugin depends on which, g...", "gated": 0, "source_path": "plugins/cwc-design/skills/architecture-wireframes/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_architecture_wireframes" }, { "id": "skl_data_shape_svg", "name": "data-shape-svg", "plugin": "cwc-design", "description": "Render a real, source-verified SVG diagram — a pipeline/system architecture diagram (every node a real function/stage that exists in the codebase, grep-verified before drawing) or a data-shape /exploration diagram (every stat a real, computed number, never invented). Generalizes the pipeline-architecture-diagram pattern first built for agenttables.com (workers/agenttables/src/ pipeline-viz.ts) into a reusable methodology. For data-shape diagrams specifically, consumes the structured JSON produced by the sibling plugin cwc-data-engineering's `parquet-shape-profiler` skill (DatasetShapeProfile: row/column counts, per-column type, null rate, numeric percentiles or categorical top-values) — this skill owns turning that structured, already-computed data into SVG markup; it does NOT compute statistics itself. Use when the user wants an architecture diagram of real code, a visualization of a dataset's shape/structure, or a \"what does the data look like\" exploration diagram. Do NOT use for the fleet's own system/plugin architecture (sibling `architecture-wireframes` skill's territory — Figma/FigJam boards, a different scope) or for computing the underlying data statistics yourself (that's cwc-data-engineering's job; this skill only visualizes numbers it's handed).", "summary": "Render a real, source-verified SVG diagram — a pipeline/system architecture diagram (every node a real function/stage that exists in the codebase, grep-verified before drawing) or a data-shape /exploration diagram (every stat a real, computed number, never invented).", "gated": 0, "source_path": "plugins/cwc-design/skills/data-shape-svg/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_data_shape_svg" }, { "id": "skl_dependency_graph_sync", "name": "dependency-graph-sync", "plugin": "cwc-design", "description": "Keep the DESIGN VIEW of the fleet's dependency graph — the subagentdevices.com/architecture SVG node-link map (hand-maintained ARCH_NODES/ARCH_EDGES_* arrays) and the Figma architecture board (wireframes.json ledger) — in sync with the GENERATED source of truth, workers/build-graph.json. Use when a cross-site fetch is added or removed, a new worker/site ships or one is retired, the user asks \"is the architecture map current\", \"check map drift\", \"arch drift\", \"sync the diagram\", \"does the map match reality\", after scripts/build-graph.py regenerates, or before publishing/screenshotting any architecture visual. Do NOT use for creating or CRUD-ing boards themselves (sibling architecture-wireframes skill), for build/deploy staleness of workers against kit changes (scripts/worker-build.sh — cwc-engineering/cwc-deploy territory), or for live health of fetch targets alone (build-graph.py --check-refs covers that without this skill).", "summary": "Keep the DESIGN VIEW of the fleet's dependency graph — the subagentdevices.com/architecture SVG node-link map (hand-maintained ARCH_NODES/ARCH_EDGES_* arrays) and the Figma architecture board (wireframes.json ledger) — in sync with the GENERATED source of truth, workers/build-graph.json.", "gated": 0, "source_path": "plugins/cwc-design/skills/dependency-graph-sync/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_dependency_graph_sync" }, { "id": "skl_design_playbook", "name": "design-playbook", "plugin": "cwc-design", "description": "Design coworker playbook — work the 8-token design system and ship Claude Hardware Buddy character packs through the gated publish loop. Use when the user wants design tokens, a token lint (WCAG contrast), a buddy character pack built from tokens, or a pack published (which requires operator approval on the buddy device). Trigger on \"design tokens\", \"lint the palette\", \"build a buddy pack\", \"publish the pack\", or follow-up design work after /design.", "summary": "Design coworker playbook — work the 8-token design system and ship Claude Hardware Buddy character packs through the gated publish loop.", "gated": 1, "source_path": "plugins/cwc-design/skills/design-playbook/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_design_playbook" }, { "id": "skl_document", "name": "document", "plugin": "cwc-design", "description": "Produce the \"Document\" project shape (one of claude.ai/design's 5 canonical shapes — Prototype/Slides/Document/Wireframe/Animation): research briefs, synthesis reports, accessibility audits, and design critiques — anything whose output IS a written document, not a visual mockup. Use when the user wants to plan/synthesize user research, run a WCAG audit, get structured design feedback on a mockup, or produce any design-related written report. Do NOT use for a live wireframe/layout artifact (sibling `wireframe` skill) or for slide decks (sibling `slides` skill).", "summary": "Produce the \"Document\" project shape (one of claude.ai/design's 5 canonical shapes — Prototype/Slides/Document/Wireframe/Animation): research briefs, synthesis reports, accessibility audits, and design critiques — anything whose output IS a written document, not a visual mockup.", "gated": 0, "source_path": "plugins/cwc-design/skills/document/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_document" }, { "id": "skl_fleet_design_system", "name": "fleet-design-system", "plugin": "cwc-design", "description": "Owns the real, deployed CSS design system shared by 66 subagent*.com Workers (workers/_shared/kit/css.ts + shell.ts), its live showcase (subagentcowork.com/design-system), its per-technique adoption evidence (subagentdevices.com), a typed/tested drift check against subagentcoworkers.com/deck, and a portable JSON export for attaching/creating this same design system on claude.ai/design or other cross-surface consumers (e.g. Claude Cowork on web). Use when the user wants \"design system\", \"design tokens\", \"fleet CSS\", to add a new site's accent, to check whether a site matches the design system, to sync the deck with the kit, or to attach/export this design system to a claude.ai/design project (Prototype/Slides/Document/ Wireframe/Animation). Do NOT use for editing workers/_shared/kit/*.ts itself (a fleet-wide hash-relock event — that's cwc-deploy/cwc-engineering territory), the Hardware Buddy 8-token character-pack system (sibling design-playbook skill), or the architecture SVG map (sibling dependency-graph-sync skill).", "summary": "Owns the real, deployed CSS design system shared by 66 subagent*.com Workers (workers/_shared/kit/css.ts + shell.ts), its live showcase (subagentcowork.com/design-system), its per-technique adoption evidence (subagentdevices.com), a typed/tested drift check against subagentcoworkers.com/deck, and...", "gated": 0, "source_path": "plugins/cwc-design/skills/fleet-design-system/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_fleet_design_system" }, { "id": "skl_ink", "name": "ink", "plugin": "cwc-design", "description": "Ink terminal renderer for json-render that turns JSON specs into interactive terminal UIs. Use when working with @json-render/ink, building terminal UIs from JSON, creating terminal component catalogs, or rendering AI-generated specs in the terminal.", "summary": "Ink terminal renderer for json-render that turns JSON specs into interactive terminal UIs.", "gated": 0, "source_path": "plugins/cwc-design/skills/ink/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_ink" }, { "id": "skl_liquid_icon", "name": "liquid-icon", "plugin": "cwc-design", "description": "Generate a macOS/iOS app icon in the fleet's Liquid-Glass / entity-graph brand style and compose it into an .icns wired into an .app bundle. Two stages: (1) generate the icon ART with the higgsfield-generate skill (the `higgsfield` CLI — free-tier models flux_2 / grok_image / flux_kontext work; gpt_image_2 / nano_banana need a paid plan), brand = black ground + cyan #51c4ff + green #7bd88f + a glowing node-link graph; (2) build the .icns deterministically via `scripts/build-icns.sh` (sips + iconutil + PlistBuddy — macOS built-ins, no deps) and set CFBundleIconFile. Use when the user asks for an app icon, a Liquid-Glass icon, a logo for a Swift app, or to (re)brand a .app's Dock icon. Trigger on \"app icon\", \"liquid glass icon\", \"logo for the app\", \"brand the app icon\", \"Icon Composer\". Do NOT use for in-app SwiftUI glass styling (that's the app's own `.glassEffect` code) or for non-icon marketing art (use higgsfield-generate直接).", "summary": "Generate a macOS/iOS app icon in the fleet's Liquid-Glass / entity-graph brand style and compose it into an .icns wired into an .app bundle.", "gated": 0, "source_path": "plugins/cwc-design/skills/liquid-icon/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_liquid_icon" }, { "id": "skl_preview_verify_deploy", "name": "preview-verify-deploy", "plugin": "cwc-design", "description": "Verify a Cloudflare Worker change in a real browser before shipping it — start the Worker via this repo's committed `.claude/launch.json` (named, persistent `wrangler dev` configs, one per Worker; not a temporary scratchpad config), reload, check console/network for errors, assert DOM state via `preview_eval` (screenshots are known-unreliable in this harness — they render solid black after any non-zero scroll, a tool-level rendering bug independently confirmed twice this repo's history), then `wrangler deploy` via Desktop Commander and curl-verify the live result. Use this any time a change touches something a Worker actually renders/serves (HTML, CSS, client-side script, an API route) — not for backend-only or non-previewable changes. Sibling to `plugins/cwc-engineering/skills/preview-layout-verify` (layout-measurement invariants only, clipping/collision/overflow) — that skill's invariant sweep is a good deep-dive once a page loads; this skill owns the outer loop (start → verify → deploy → confirm) and the two compose: run this skill's loop, and if the change is layout-shaped, drop into that skill's invariant sweep during the \"verify\" step.", "summary": "Verify a Cloudflare Worker change in a real browser before shipping it — start the Worker via this repo's committed `.claude/launch.json` (named, persistent `wrangler dev` configs, one per Worker; not a temporary scratchpad config), reload, check console/network for errors, assert DOM state via `...", "gated": 0, "source_path": "plugins/cwc-design/skills/preview-verify-deploy/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_preview_verify_deploy" }, { "id": "skl_prototype", "name": "prototype", "plugin": "cwc-design", "description": "Produce the \"Prototype\" project shape (one of claude.ai/design's 5 canonical shapes — Prototype/Slides/Document/Wireframe/Animation): a high-fidelity, interactive artifact, plus the UX microcopy that makes it feel real (CTAs, error messages, empty states, confirmation dialogs). Use when the user wants a working interactive mockup, needs copy for a specific interface moment, or wants to see the fleet design system's own live interactive reference (subagentcowork.com/design-system's accent picker). Do NOT use for a static layout without interactivity (sibling `wireframe` skill) or for a written report (sibling `document` skill).", "summary": "Produce the \"Prototype\" project shape (one of claude.ai/design's 5 canonical shapes — Prototype/Slides/Document/Wireframe/Animation): a high-fidelity, interactive artifact, plus the UX microcopy that makes it feel real (CTAs, error messages, empty states, confirmation dialogs).", "gated": 0, "source_path": "plugins/cwc-design/skills/prototype/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_prototype" }, { "id": "skl_session_record_html", "name": "session-record-html", "plugin": "cwc-design", "description": "Scaffold a standalone HTML session-record file, in the family's shared design tokens, so any coworker or agent can leave a durable, visual record of what it completed in a session — for human recall, not machine consumption. Use when the user says \"record this session\", \"write up what we did\", \"create a session record\", \"document this session as an artifact\", or asks for a visual summary of completed work to look back on later. Do NOT use for the live coworkers-console dashboard itself (index.html + serve.py — a different, API-backed artifact), and do NOT fabricate content — every fact in the record must be something the session actually verified, not a plausible-sounding summary.", "summary": "Scaffold a standalone HTML session-record file, in the family's shared design tokens, so any coworker or agent can leave a durable, visual record of what it completed in a session — for human recall, not machine consumption.", "gated": 0, "source_path": "plugins/cwc-design/skills/session-record-html/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_session_record_html" }, { "id": "skl_slides", "name": "slides", "plugin": "cwc-design", "description": "Produce the \"Slides\" project shape (one of claude.ai/design's 5 canonical shapes — Prototype/Slides/Document/Wireframe/Animation), grounded in the one real slide deck already in this repo family — subagentcoworkers.com/deck, a real 12-slide investor deck with a semantic per-slide accent system. Use when the user wants to build a pitch/investor deck, a slide-based presentation, or wants a new deck to follow the same real conventions (eyebrow labels, per-slide accent, slide-number counter, keyboard navigation). Do NOT use for a single static document (sibling `document` skill) or for the deck's own token block maintenance (that's `fleet-design-system`'s drift-check territory — see below).", "summary": "Produce the \"Slides\" project shape (one of claude.ai/design's 5 canonical shapes — Prototype/Slides/Document/Wireframe/Animation), grounded in the one real slide deck already in this repo family — subagentcoworkers.com/deck, a real 12-slide investor deck with a semantic per-slide accent system.", "gated": 0, "source_path": "plugins/cwc-design/skills/slides/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_slides" }, { "id": "skl_tdd_red_green_refactor", "name": "tdd-red-green-refactor", "plugin": "cwc-design", "description": "Enforces a disciplined Red-Green-Refactor (TDD) workflow in TypeScript/Node.js. Use this whenever creating new features, fixing bugs, or migrating logic to ensure high-quality, verifiable implementations.", "summary": "Enforces a disciplined Red-Green-Refactor (TDD) workflow in TypeScript/Node.js.", "gated": 0, "source_path": "plugins/cwc-design/skills/tdd/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_tdd_red_green_refactor" }, { "id": "skl_typed_service_contracts", "name": "typed-service-contracts", "plugin": "cwc-design", "description": "Architecture standard for building robust, type-safe TypeScript services using the \"Spec and Handler\" pattern. Use when building CLIs, libraries, or complex business logic.", "summary": "Architecture standard for building robust, type-safe TypeScript services using the \"Spec and Handler\" pattern.", "gated": 0, "source_path": "plugins/cwc-design/skills/typed-service-contracts/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_typed_service_contracts" }, { "id": "skl_wireframe", "name": "wireframe", "plugin": "cwc-design", "description": "Produce the \"Wireframe\" project shape (one of claude.ai/design's 5 canonical shapes — Prototype/Slides/Document/Wireframe/Animation): layout + component structure before visual polish. Three modes — audit an existing layout for token/naming consistency, document a component's variants/states/props, or extend the system with a new pattern — plus a full developer-handoff spec (measurements, states, breakpoints, edge cases) once a wireframe is ready to leave design. Use when the user wants to wireframe a screen, audit a design system for consistency, document a component, design a new pattern, or generate a handoff spec. Do NOT use for high-fidelity interactive mockups (sibling `prototype` skill) or for editing `workers/_shared/kit/*.ts` itself (sibling `fleet-design-system` skill's territory).", "summary": "Produce the \"Wireframe\" project shape (one of claude.ai/design's 5 canonical shapes — Prototype/Slides/Document/Wireframe/Animation): layout + component structure before visual polish.", "gated": 0, "source_path": "plugins/cwc-design/skills/wireframe/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_wireframe" }, { "id": "skl_agentcrawls_worker", "name": "agentcrawls-worker", "plugin": "cwc-engineering", "description": "The agentcrawls.com Python Worker (R2 + Cache API + Secrets Store + Cron Trigger) — build/ship boundary, secret name/scope, and how it relates to the local scrapyd/scrapydweb crawler. Use on \"agentcrawls Worker\", \"deploy agentcrawls\", or /agentcrawls-worker.", "summary": "The agentcrawls.com Python Worker (R2 + Cache API + Secrets Store + Cron Trigger) — build/ship boundary, secret name/scope, and how it relates to the local scrapyd/scrapydweb crawler.", "gated": 0, "source_path": "plugins/cwc-engineering/skills/agentcrawls-worker/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_agentcrawls_worker" }, { "id": "skl_atlas_graph_prototype", "name": "atlas-graph-prototype", "plugin": "cwc-engineering", "description": "No-JVM fallback prototype of the Netflix Atlas Graph API (`GET /api/v1/graph?q=[&OPTIONS]`, port 7102 — real Atlas owns :7101 since the 2026-07-05 swap) over the Coworkers board's telemetry.jsonl — stdlib ThreadingHTTPServer + matplotlib Agg (always headless), uv-managed (`uv run pytest` / `uv run atlas-graph serve|render`). Implements the param subset pinned from the mirrored graph.md (q, s, e, tz, w, h, title, step, format=png|json, no_legend, stack — everything else 400s as deferred) and the ASL RPN subset the `requires: subset` presets in schemas/atlas-graph-presets.json need (:eq/:and/:sum/:by; real-atlas-only presets like completed-anomaly-des are skipped). Bucketing per time-series.md: even step boundaries, floor normalization, day-granular buckets honor tz; values are per-bucket COUNTS (honest rate-vs-count departure documented in README). Use to serve or render board telemetry graphs without the JVM, add a graph the Swift AtlasGraphView needs, or extend the param/ASL subset (same URLs — the contract is the product). Trigger on \"atlas graph\", \"graph API\", \"telemetry graph\", \"render the board's PNG\", \"completed per day\", \"blocked timeline\", \"7102\", \"no-JVM fallback\". Do NOT use for the Swift AtlasGraphDef/View (cwc-swift-engineering), the telemetry star schema (cwc-data-engineering's session-telemetry-erd), or running real JVM Atlas (scripts/atlas/run-atlas.sh on :7101).", "summary": "No-JVM fallback prototype of the Netflix Atlas Graph API (`GET /api/v1/graph?q=[&OPTIONS]`, port 7102 — real Atlas owns :7101 since the 2026-07-05 swap) over the Coworkers board's telemetry.jsonl — stdlib ThreadingHTTPServer + matplotlib Agg (always headless), uv-managed (`uv run pytest` /...", "gated": 0, "source_path": "plugins/cwc-engineering/skills/atlas-graph-prototype/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_atlas_graph_prototype" }, { "id": "skl_code_mode_ptc", "name": "code-mode-ptc", "plugin": "cwc-engineering", "description": "Token-thrift execution doctrine: prefer CLIs and python scripts over resident plugins/tools, batch tool work programmatically (PTC/code-mode), route heavy work to sonnet-tier subagents, and answer lookups from caches (ragcache) — so context carries decisions, not tooling. Use when choosing HOW to execute any multi-step task, when tempted to install a project-level plugin, or on \"code mode\", \"PTC\", \"minimize tokens\".", "summary": "Token-thrift execution doctrine: prefer CLIs and python scripts over resident plugins/tools, batch tool work programmatically (PTC/code-mode), route heavy work to sonnet-tier subagents, and answer lookups from caches (ragcache) — so context carries decisions, not tooling.", "gated": 0, "source_path": "plugins/cwc-engineering/skills/code-mode-ptc/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_code_mode_ptc" }, { "id": "skl_code_review_workflow", "name": "code-review-workflow", "plugin": "cwc-engineering", "description": "Grounds this repo's use of Claude Code's real Code Review capabilities: the managed, billed GitHub-PR-review service (admin-enabled per org, posts inline 🔴/🟡/🟣-severity comments, customizable via REVIEW.md) and the local, unbilled-beyond-usage /code-review terminal command (--comment, --fix, effort levels, /code-review ultra for a deeper cloud pass). Use before claiming a PR was \"reviewed by Claude\" — that phrase means one of two specific, different things, and this skill tells you which commands actually invoke which. Use when asked to tune review behavior — that's REVIEW.md, not a custom review script.", "summary": "Grounds this repo's use of Claude Code's real Code Review capabilities: the managed, billed GitHub-PR-review service (admin-enabled per org, posts inline 🔴/🟡/🟣-severity comments, customizable via REVIEW.md) and the local, unbilled-beyond-usage /code-review terminal command (--comment, --fix, effo...", "gated": 0, "source_path": "plugins/cwc-engineering/skills/code-review-workflow/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_code_review_workflow" }, { "id": "skl_docker_mcp_build", "name": "docker-mcp-build", "plugin": "cwc-engineering", "description": "Deterministic, token-free \"build\" gate for the Docker MCP Toolkit profile mirror (skills/docker-mcp-config/mcp_config.yaml). Use when asked to build, validate, lock, or version the MCP toolkit profiles, wire the pre-commit gate, bump the mirror schema version, or investigate a stale mcp_config.lock.json / blocked commit. Pydantic + SemVer + managed enums; runs at pre-commit and in make — never spends Claude tokens.", "summary": "Deterministic, token-free \"build\" gate for the Docker MCP Toolkit profile mirror (skills/docker-mcp-config/mcp_config.yaml).", "gated": 0, "source_path": "plugins/cwc-engineering/skills/docker-mcp-build/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_docker_mcp_build" }, { "id": "skl_docker_mcp_config", "name": "docker-mcp-config", "plugin": "cwc-engineering", "description": "Type-safe, version-controlled mirror of the Docker MCP Toolkit profiles on this Mac (mcp_config.yaml: top-level keys = profile ids, values = enabled servers + config + ${VAR} secrets). Use when asked to sync/mirror MCP toolkit profiles, add a coworker MCP profile, enable/disable a connector for a coworker, fix the redis MCP server, or wire a new Claude client to a profile. Diff-first: validate.py (offline gate) -> sync.py (dry-run diff) -> operator-gated sync.py --apply. Toolkit state is edited ONLY via the docker mcp CLI driven by this skill — never the SQLite at ~/.docker/mcp/mcp-toolkit.db; any Docker Desktop UI edit must be re-mirrored here immediately.", "summary": "Type-safe, version-controlled mirror of the Docker MCP Toolkit profiles on this Mac (mcp_config.yaml: top-level keys = profile ids, values = enabled servers + config + ${VAR} secrets).", "gated": 0, "source_path": "plugins/cwc-engineering/skills/docker-mcp-config/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_docker_mcp_config" }, { "id": "skl_docs_rag", "name": "docs-rag", "plugin": "cwc-engineering", "description": "Token-thrift RAG over the docs/ mirrors: three-tier cache (functools.lru_cache → redis → sqlite FTS5) at crawlers/ragcache.py. USE THIS FIRST whenever a session needs \"what do the mirrors say about X\" — query for snippets (~200 tokens), read the one file that matters, never bulk-read mirror trees into context. Triggers: \"search the docs\", \"what did we crawl about\", \"find in mirrors\", /docs-rag.", "summary": "Token-thrift RAG over the docs/ mirrors: three-tier cache (functools.lru_cache → redis → sqlite FTS5) at crawlers/ragcache.py.", "gated": 0, "source_path": "plugins/cwc-engineering/skills/docs-rag/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_docs_rag" }, { "id": "skl_durable_mac_toolchain", "name": "durable-mac-toolchain", "plugin": "cwc-engineering", "description": "Doctor + install + autoresolve for the developer Mac (darwin/zsh/Homebrew) in ONE skill — the darwin member of the durable-toolchain family (doctor/install/autoresolve), combined and scoped to this workstation with a typed manifest. Use whenever a tool call fails with a missing-binary/PATH/TLS/zsh error (\"No Chrome binary for devtools\", \"command not found\", \"CERTIFICATE_VERIFY_FAILED\", \"(eval): no matches found\"), before browser-verification work, or on \"doctor\", \"install tools\", \"fix my Mac toolchain\", \"/durable-mac-toolchain\". Never free-recall an install command — copy validator output verbatim.", "summary": "Doctor + install + autoresolve for the developer Mac (darwin/zsh/Homebrew) in ONE skill — the darwin member of the durable-toolchain family (doctor/install/autoresolve), combined and scoped to this workstation with a typed manifest.", "gated": 0, "source_path": "plugins/cwc-engineering/skills/durable-mac-toolchain/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_durable_mac_toolchain" }, { "id": "skl_engineering_playbook", "name": "engineering-playbook", "plugin": "cwc-engineering", "description": "Engineering playbook for the subagent*.com fleet. Use when the user asks to \"add a new worker\", \"scaffold a new subagent site\", \"build a new primitive\", \"extend the API on \", \"add a route/column to \", \"regenerate the agents\", \"run agent-gen\", \"keep the agent schema mirrors in sync\", \"check schema drift\" between agent.rs / schema.ts / schema.py, or asks how this repo's typed generated-agent system works. Do NOT use for executing a deploy — cwc-deploy owns the ship step (install/typecheck/wrangler/secrets/D1-migrate on the real Mac); this skill builds the code that cwc-deploy then ships. Do NOT use for warehouse analytics (cwc-data) or fleet-wide audits/health sweeps (cwc-operations).", "summary": "Engineering playbook for the subagent*.com fleet.", "gated": 0, "source_path": "plugins/cwc-engineering/skills/engineering-playbook/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_engineering_playbook" }, { "id": "skl_git_github_integration", "name": "git-github-integration", "plugin": "cwc-engineering", "description": "Typed catalog of git/GitHub integration surfaces for an operator with NO ANTHROPIC_API_KEY — claude.ai 20x (Max-tier) subscription via OAuth in Claude Desktop only. Covers gh CLI (already authenticated), the official commit-commands plugin, the hosted GitHub MCP server, GitHub Actions (including the real claude_code_oauth_token path via `claude setup-token`, verified against the action's own action.yml, not just the prose docs), and GitHub Enterprise Server as a documented-but-inapplicable section. Use on \"set up GitHub integration\", \"commit and push\", \"add a GitHub Action\", \"wire the GitHub MCP\", or before assuming ANTHROPIC_API_KEY is required for any GitHub-adjacent Claude Code feature — check integration.json first, it usually isn't.", "summary": "Typed catalog of git/GitHub integration surfaces for an operator with NO ANTHROPIC_API_KEY — claude.ai 20x (Max-tier) subscription via OAuth in Claude Desktop only.", "gated": 0, "source_path": "plugins/cwc-engineering/skills/git-github-integration/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_git_github_integration" }, { "id": "skl_model_routing_hooks", "name": "model-routing-hooks", "plugin": "cwc-engineering", "description": "The zero-token model router + telemetry spine for Agent/Task dispatches: four stdlib-only python3 hooks in scripts/hooks/ (PreToolUse router rewriting an absent `model` param to sonnet via updatedInput, async PostToolUse usage capture, SessionStart policy context, async Stop flush of redis route:queue into postgres agent_dispatch_events). Use when the user asks \"why did my subagent run on sonnet\", \"route subagents to a cheaper model\", \"model router\", \"routing-rules.json\", \"route:queue\", \"agent_dispatch_events\", \"subagent token telemetry\", \"pin a subagent's model\", or wants to change/disable the routing policy or query per-dispatch token spend. This skill owns the hook layer only — the eval and optimizer that consume agent_dispatch_events live at eval/ and optimizer/ inside this same skill dir (built separately; reference by path).", "summary": "The zero-token model router + telemetry spine for Agent/Task dispatches: four stdlib-only python3 hooks in scripts/hooks/ (PreToolUse router rewriting an absent `model` param to sonnet via updatedInput, async PostToolUse usage capture, SessionStart policy context, async Stop flush of redis route:...", "gated": 0, "source_path": "plugins/cwc-engineering/skills/model-routing-hooks/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_model_routing_hooks" }, { "id": "skl_platform_doctor", "name": "platform-doctor", "plugin": "cwc-engineering", "description": "Doctor + version-drift reporter + gated installer for Claude-brand applications on this Mac (macOS 27.0 / arm64) — Claude Desktop, Claude Bioscience, Claude Code CLI, the VS Code extension, JetBrains Toolbox, and Chrome (chrome-devtools MCP dependency). `validate.py --doctor` reports PRESENT/MISSING plus the installed version string; `pincheck.py --check` compares installed vs. this repo's pinned version (CLAUDE.md) vs. the real live \"latest\" (GitHub releases / VS Code Marketplace API) where a verifiable source exists; `resolve.py --plan| --resolve` runs a fail-fast, logging, gated installer per app. Use on \"what's installed\", \"check claude apps\", \"is claude code out of date\", \"install claude bioscience\", \"platform doctor\", \"/platform-doctor\", before recommending an install, or whenever a Claude-app-related tool call fails with a missing-app error. Never free-recall an app path, version command, \"latest version\" number, or install command — copy validate.py/pincheck.py/resolve.py output verbatim; an app with no verifiable latest-version source is reported UNKNOWN, never guessed; a gated install is ESCALATED with the exact command, never auto-run.", "summary": "Doctor + version-drift reporter + gated installer for Claude-brand applications on this Mac (macOS 27.0 / arm64) — Claude Desktop, Claude Bioscience, Claude Code CLI, the VS Code extension, JetBrains Toolbox, and Chrome (chrome-devtools MCP dependency).", "gated": 0, "source_path": "plugins/cwc-engineering/skills/platform-doctor/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_platform_doctor" }, { "id": "skl_preview_layout_verify", "name": "preview-layout-verify", "plugin": "cwc-engineering", "description": "Verify layout/CSS changes with a real browser via the Claude Preview harness (mcp__Claude_Preview__*) — measurements first, screenshots second. Use after any deck/site layout change, when the operator reports sizing/overlap/wrapping bugs, or when chrome-devtools MCP is unavailable (no Chrome binary). Proven on the Launchpad deck v9 fix (2026-07-03): clipping/collision/overflow found and eliminated by numbers, not eyeballs.", "summary": "Verify layout/CSS changes with a real browser via the Claude Preview harness (mcp__Claude_Preview__*) — measurements first, screenshots second.", "gated": 0, "source_path": "plugins/cwc-engineering/skills/preview-layout-verify/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_preview_layout_verify" }, { "id": "skl_scrapy_docs_crawler", "name": "scrapy-docs-crawler", "plugin": "cwc-engineering", "description": "Keep the docs/ mirrors fresh with the Scrapy sitemap crawler (crawlers/): md-first fetching (.md variant tried per-URL, deterministic HTML distiller as fallback), bloom-filter cross-run URL skipping, and sha256 CDC so unchanged pages never rewrite. Use to refresh claude.com/anthropic.com/support/platform/code/scrapyd/scrapy mirrors, on \"crawl the docs\", \"update mirrors\", \"refresh claude.com blog\", or /scrapy-docs-crawler. Small-scale by design — sitemaps only, polite throttle, hard page caps.", "summary": "Keep the docs/ mirrors fresh with the Scrapy sitemap crawler (crawlers/): md-first fetching (.md variant tried per-URL, deterministic HTML distiller as fallback), bloom-filter cross-run URL skipping, and sha256 CDC so unchanged pages never rewrite.", "gated": 0, "source_path": "plugins/cwc-engineering/skills/scrapy-docs-crawler/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_scrapy_docs_crawler" }, { "id": "skl_scrapyd_scrapydweb_local", "name": "scrapyd-scrapydweb-local", "plugin": "cwc-engineering", "description": "Run the local scrapyd job daemon + scrapydweb visual dashboard for the docs crawler (crawlers/claude_docs) — the manual/full-recrawl half of crawling, distinct from the daily automated Cloudflare Cron Trigger on workers/agentcrawls. Use on \"start the crawl dashboard\", \"run scrapydweb\", \"trigger a full re-crawl visually\", or /scrapyd-scrapydweb-local.", "summary": "Run the local scrapyd job daemon + scrapydweb visual dashboard for the docs crawler (crawlers/claude_docs) — the manual/full-recrawl half of crawling, distinct from the daily automated Cloudflare Cron Trigger on workers/agentcrawls.", "gated": 0, "source_path": "plugins/cwc-engineering/skills/scrapyd-scrapydweb-local/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_scrapyd_scrapydweb_local" }, { "id": "skl_security_guidance_wiring", "name": "security-guidance-wiring", "plugin": "cwc-engineering", "description": "Grounds this repo's use of Anthropic's official security-guidance plugin (three review layers: per-edit pattern match, end-of-turn model-backed diff review, commit/push agentic review) and the two files that extend it — .claude/claude-security-guidance.md (freeform guidance) and .claude/security-patterns.yaml (custom per-edit regex/substring rules). Use before writing a custom security-scanning script for this repo — the official plugin already does per-edit, per-turn, and per-commit review built on the same hooks mechanism documented in hooks-guide.md; extend it via its own config files instead of duplicating its logic.", "summary": "Grounds this repo's use of Anthropic's official security-guidance plugin (three review layers: per-edit pattern match, end-of-turn model-backed diff review, commit/push agentic review) and the two files that extend it — .claude/claude-security-guidance.md (freeform guidance) and .claude/security-...", "gated": 0, "source_path": "plugins/cwc-engineering/skills/security-guidance-wiring/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_security_guidance_wiring" }, { "id": "skl_system_data_doctor", "name": "system-data-doctor", "plugin": "cwc-engineering", "description": "Measure, rank, and (gated) clean the macOS \"System Data\" storage category on this Mac — the AI-tooling logs, VM disks, hidden dotfile stores, and caches that repeatedly fill the disk. Use when the disk is full or filling, when asked to free up space, investigate System Data, find what ate the disk, or schedule storage hygiene. Evidence first: --doctor is a read-only ranked scan with a dated trend report; cleaning is manifest-scoped and gate-controlled.", "summary": "Measure, rank, and (gated) clean the macOS \"System Data\" storage category on this Mac — the AI-tooling logs, VM disks, hidden dotfile stores, and caches that repeatedly fill the disk.", "gated": 0, "source_path": "plugins/cwc-engineering/skills/system-data-doctor/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_system_data_doctor" }, { "id": "skl_worker_build", "name": "worker-build", "plugin": "cwc-engineering", "description": "The engineering view of building ONE Cloudflare Worker correctly in the 71-worker fleet: scripts/worker-build.sh (kit-dep resolution → tsc → drift gate vs .build-lock.json) and scripts/build-graph.py (the committed workers/build-graph.json dependency graph). Use when the user says \"build a worker\", \"typecheck \", \"kit drift\", \"stale workers\", \"who imports shell.ts\", \"who fetches subagentcontracts.com\", \"blast radius of a kit edit\", \"add a kit import\", \"regenerate the build graph\", \".build-lock.json\", or after any edit under workers/_shared/kit/ or workers//src/. This skill builds and gates; it does NOT ship — deploy execution is cwc-deploy's deploy-playbook. Do NOT use for the fleet-wide sweep after a kit edit lands (cwc-operations:dependency-drift-audit) or for approving a new runtime dependency between sites (cwc-product-management:dependency-contracts).", "summary": "The engineering view of building ONE Cloudflare Worker correctly in the 71-worker fleet: scripts/worker-build.sh (kit-dep resolution → tsc → drift gate vs .build-lock.json) and scripts/build-graph.py (the committed workers/build-graph.json dependency graph).", "gated": 0, "source_path": "plugins/cwc-engineering/skills/worker-build/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_worker_build" }, { "id": "skl_worker_conventions_lint", "name": "worker-conventions-lint", "plugin": "cwc-engineering", "description": "Structural lint for a workers// directory against the fleet's five real conventions (each grep-verified 2026-07-09): direct-path ../../_shared/kit/ imports (barrel imports evade scripts/worker-build.sh's kit drift gate), account_id pinned to the fleet Cloudflare account in wrangler.toml, an apex+www [[routes]] pair, [observability] enabled = true, and the package.json allowScripts block. Pure functions over file contents plus a report-only CLI (node scripts/lint.ts ). Use when scaffolding a new worker from workers/_template, reviewing a worker PR, or when a wrangler deploy silently exits 1 / esbuild-workerd binaries silently fail to install and you want the structural cause named. Do NOT use for kit HASH drift — scripts/worker-build.sh gates kit hashes; this lints STRUCTURE and complements, never replaces, that gate — nor for CSS/design-token conventions (cwc-design's fleet-design-system skill) or for executing deploys (cwc-deploy).", "summary": "Structural lint for a workers// directory against the fleet's five real conventions (each grep-verified 2026-07-09): direct-path ../../_shared/kit/ imports (barrel imports evade scripts/worker-build.sh's kit drift gate), account_id pinned to the fleet Cloudflare account in wrangler.tom...", "gated": 0, "source_path": "plugins/cwc-engineering/skills/worker-conventions-lint/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_worker_conventions_lint" }, { "id": "skl_cost_revenue_monthly", "name": "cost-revenue-monthly", "plugin": "cwc-finance", "description": "The monthly money picture for this solo-founder AI startup, in one pass: fixed costs vs the two-line-item baseline, usage signals read (never invented) from existing provenance-tagged snapshots, a revenue-readiness checklist (revenue is $0 so far; the path is Cloudflare Monetization Gateway + x402 + Stripe when it ships), and one written report. Use when the user asks for a \"monthly close\", \"cost report\", \"burn report\", \"revenue readiness\", \"monthly finance pass\", or \"close the month\". Do NOT use to approve a new spend (sibling spend-tracking owns the spend_commit gate), for the Monetization Gateway rollout details (sibling monetization-gateway), or to run ad-hoc GraphQL Analytics queries (cwc-analytics-engineering's analytics-playbook owns that discipline — this skill only reads what's already recorded).", "summary": "The monthly money picture for this solo-founder AI startup, in one pass: fixed costs vs the two-line-item baseline, usage signals read (never invented) from existing provenance-tagged snapshots, a revenue-readiness checklist (revenue is $0 so far; the path is Cloudflare Monetization Gateway + x40...", "gated": 0, "source_path": "plugins/cwc-finance/skills/cost-revenue-monthly/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_cost_revenue_monthly" }, { "id": "skl_credit_realizable_value", "name": "credit-realizable-value", "plugin": "cwc-finance", "description": "Typed, pure-function calculator that makes the operator's standing rule executable: \"vendor credits at realizable value, never face value\" (plugins/cwc-finance/memory/v2/directives/ core.md, verbatim). A CreditGrant {vendor, face_value, expiry_months, monthly_cap, eligible_spend_forecast_monthly} resolves to realizable_value = min over the binding constraints (face ceiling, cap x expiry window, honest usage forecast x expiry window), with a per-constraint explanation string so the output SHOWS which term binds, plus a booking-line renderer (\"book min(credit, real usage)\"). Use when the user asks \"what's this credit actually worth\", \"book this vendor credit\", \"realizable value\", \"credit reality check\", or before any credit number enters a ledger, plan, or application. Do NOT use to approve the spend the credit offsets (sibling spend-tracking owns the spend_commit gate), to fetch or guess a vendor's terms (per the rule, no credit number enters the ledger without its terms page fetched and dated — this calculator consumes those terms, it never invents them), or as a revenue signal (startup-finance: \"monetization-ready\" is not revenue).", "summary": "Typed, pure-function calculator that makes the operator's standing rule executable: \"vendor credits at realizable value, never face value\" (plugins/cwc-finance/memory/v2/directives/ core.md, verbatim).", "gated": 0, "source_path": "plugins/cwc-finance/skills/credit-realizable-value/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_credit_realizable_value" }, { "id": "skl_excel_toolchain", "name": "excel-toolchain", "plugin": "cwc-finance", "description": "Durable, autoresolving Excel toolchain for cwc-finance on the developer Mac (darwin-arm64) — typed manifest (toolchain.json) + validate.py doctor/resolve, modeled on cwc-engineering's durable-mac-toolchain. Use on \"excel toolchain\", \"fix excel setup\", \"xlwings\", \"recalc setup\", when an Excel script fails with \"command not found: uv\", an import error on openpyxl/xlwings/pydantic/yaml, or AppleScript error -1743, and as step 0 of EVERY cwc-finance Excel script (validate.py --resolve, programmatic). Never free-recall a check or install command — copy validator output verbatim.", "summary": "Durable, autoresolving Excel toolchain for cwc-finance on the developer Mac (darwin-arm64) — typed manifest (toolchain.json) + validate.py doctor/resolve, modeled on cwc-engineering's durable-mac-toolchain.", "gated": 0, "source_path": "plugins/cwc-finance/skills/excel-toolchain/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_excel_toolchain" }, { "id": "skl_monetization_gateway", "name": "monetization-gateway", "plugin": "cwc-finance", "description": "Content-monetization playbook for the subagent*.com fleet, grounded in Cloudflare's Monetization Gateway (announced 2026-07-01, early-access waitlist) with x402 settlement in USDC and Open USD (OUSD). Use when the user asks \"monetize the fleet\", \"monetization gateway\", \"pay per crawl\", \"x402\", \"OUSD\", \"open usd\", \"charge AI agents for content\", \"pricing rules\", or wants to know what's real vs illustrative in this repo's payment endpoints. Do NOT use to build custom payment processing in a Worker — the gateway enforces pricing at the zone edge, ahead of the Worker, and none of this family's sites process real payments today.", "summary": "Content-monetization playbook for the subagent*.com fleet, grounded in Cloudflare's Monetization Gateway (announced 2026-07-01, early-access waitlist) with x402 settlement in USDC and Open USD (OUSD).", "gated": 0, "source_path": "plugins/cwc-finance/skills/monetization-gateway/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_monetization_gateway" }, { "id": "skl_spend_tracking", "name": "spend-tracking", "plugin": "cwc-finance", "description": "Spend-tracking playbook for the subagent*.com fleet. Use when the user asks \"track this cost\", \"approve spend\", \"new vendor\", \"what are we spending\", \"record a cost observation\", \"spend ledger\", \"budget check\", \"is this spend approved\", or before provisioning anything with a price tag (a paid API tier, a paid beta, a domain purchase, a new SaaS vendor). Do NOT use for monetization/revenue strategy — that is the sibling monetization-gateway skill — and do NOT use for Anthropic-billing conceptual questions in the abstract; subagentcredits.com catalogs those concepts (deliberately with zero dollar figures) and this skill only points at it.", "summary": "Spend-tracking playbook for the subagent*.com fleet.", "gated": 0, "source_path": "plugins/cwc-finance/skills/spend-tracking/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_spend_tracking" }, { "id": "skl_startup_finance", "name": "startup-finance", "plugin": "cwc-finance", "description": "Finance for THIS startup, kept stupid simple: a two-line-item cost ledger (Cloudflare account + one $100/mo claude.ai subscription), a hard no-API-key-spend rule, monthly cost narrative, and the finance half of a Cloudflare Workers Launchpad application. Use when the user asks \"what do we spend\", \"costs\", \"runway\", \"cost ledger\", \"monthly finance review\", \"launchpad application\", \"startup programs\", \"can we afford X\", or a coworker proposes anything with a price tag. Do NOT use for the Monetization Gateway revenue playbook (sibling monetization-gateway skill) or for provenance-tagged analytics snapshots (cwc-analytics-engineering).", "summary": "Finance for THIS startup, kept stupid simple: a two-line-item cost ledger (Cloudflare account + one $100/mo claude.ai subscription), a hard no-API-key-spend rule, monthly cost narrative, and the finance half of a Cloudflare Workers Launchpad application.", "gated": 0, "source_path": "plugins/cwc-finance/skills/startup-finance/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_startup_finance" }, { "id": "skl_wbr_metrics_excel", "name": "wbr-metrics-excel", "plugin": "cwc-finance", "description": "The coworker Weekly Business Review workbook: a 21-sheet Excel deliverable (Fleet + 20 per-coworker sheets) built from the metric registry and the rpt_wbr warehouse view, with WoW deltas shipped as live Excel FORMULAS never hardcoded values. Three files: metrics_seed.py (THE metric registry — 21 Metric pydantic rows, the only writer of dim_metric), generate_wbr.py (the openpyxl workbook builder, step-0 excel-toolchain autoresolve, best- effort xlwings recalc), and this skill's own validate.py (registry gate + informational xlsx presence check) and audit_xlsx.py (the workbook correctness gate — sheet/band/formula/footnote/owner-row checks plus gate G5 warehouse reconciliation). Use on \"wbr excel\", \"generate the workbook\", \"metric registry\", \"audit the workbook\", \"seed metrics\", \"weekly business review\", \"fleet north-star metrics\", or when a WBR deliverable needs to go to the operator/PM for Monday/Wednesday review. Do NOT use for Cloudflare cache/CDN snapshots (cwc-analytics-engineering) or the Coworkers sprint board itself (cwc-data-engineering's session-telemetry-erd) — this skill owns the metric DEFINITIONS and the Excel deliverable, not the underlying fact tables' DDL or the board UI.", "summary": "The coworker Weekly Business Review workbook: a 21-sheet Excel deliverable (Fleet + 20 per-coworker sheets) built from the metric registry and the rpt_wbr warehouse view, with WoW deltas shipped as live Excel FORMULAS never hardcoded values.", "gated": 0, "source_path": "plugins/cwc-finance/skills/wbr-metrics-excel/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_wbr_metrics_excel" }, { "id": "skl_apple_support_docs", "name": "apple-support-docs", "plugin": "cwc-hardware-engineering", "description": "Query the mirrored Apple Support repair manuals — MacBook Pro (14-inch, M5) and iPhone 16 Pro — as grounding/citation material, and re-run the crawler that keeps the mirror fresh. Docs live at `docs/docs/apple/support/en-us/ /index.md` (62 files, CommonMark+GFM, frontmatter-tagged with `manuals:` and a sha256 `mirrored:` date); the crawler that produced them is `crawlers/apple-docs` (TypeScript + cheerio, `npx tsx src/cli.ts`). Use when the user asks a repair/troubleshooting question about the M5 MacBook Pro or iPhone 16 Pro, wants a citation to an Apple repair-manual procedure/screw/ part/tool, or wants the mirror re-crawled/refreshed/verified idempotent. Trigger on \"repair manual\", \"Apple support doc\", \"M5 MacBook repair\", \"iPhone 16 Pro repair\", \"re-crawl apple docs\", \"refresh apple support\", \"screw list\", \"exploded view\", \"orderable parts\". Do NOT use for live device probing (claude-app-artifacts) or network capture (network-capture) — this skill is the **read/citation** layer over static repair-manual text.", "summary": "Query the mirrored Apple Support repair manuals — MacBook Pro (14-inch, M5) and iPhone 16 Pro — as grounding/citation material, and re-run the crawler that keeps the mirror fresh.", "gated": 0, "source_path": "plugins/cwc-hardware-engineering/skills/apple-support-docs/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_apple_support_docs" }, { "id": "skl_claude_app_artifacts", "name": "claude-app-artifacts", "plugin": "cwc-hardware-engineering", "description": "Copycat-first map of what the Claude desktop app already records on-disk about devices, sessions, and the Cowork VM — plus the canonical macOS hardware/ device-management probes enterprises use. Read the real artifacts under `~/Library/Application Support/Claude/` (bridge-state, config, claude_desktop_ config, git-worktrees, ant-device-registry) and the real probes (sysctl, devicectl, profiles, system_profiler); model their SCHEMAS, never their secret values. Use when the user asks how the Claude apps track devices/sessions, what the Cowork VM bridge records, how Claude-in-Chrome pairs a device, what canonical Apple device-data surfaces exist, or wants to trace/copycat that groundwork. This is the read/trace foundation to layer on later. Trigger on \"device data\", \"Application Support\", \"bridge-state\", \"cowork vm\", \"chrome pairing\", \"device registry\", \"MDM\", \"configuration profile\", \"system_profiler\". Do NOT use for Swift code (cwc-swift-engineering) or the Xcode toolchain (cwc-xcode-engineering).", "summary": "Copycat-first map of what the Claude desktop app already records on-disk about devices, sessions, and the Cowork VM — plus the canonical macOS hardware/ device-management probes enterprises use.", "gated": 0, "source_path": "plugins/cwc-hardware-engineering/skills/claude-app-artifacts/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_claude_app_artifacts" }, { "id": "skl_mac_fleet_preflight", "name": "mac-fleet-preflight", "plugin": "cwc-hardware-engineering", "description": "One read-only preflight for THIS Mac before any fleet work — seven checks, one JSON object out, each check grounded in a real incident this repo already paid for: disk free (the 2026-07-01/2026-07-03 100%-full incidents; Xcode-beta artifacts were the culprit), the subagentjobs-postgres + subagentjobs-redis containers, redis PONG, psql SELECT 1 as the `wbr` role (the `subagentjobs` role does NOT exist locally), wrangler auth (cheap command -v by default, --wrangler-live for a real whoami), and the port-53 UDP conflict (Apple container binds mDNSResponder to 53 before Cloudflare One Client starts). Exit 0 only when the required checks (docker/redis/psql) pass; disk/wrangler/port53 are warn-only. Use before deploy waves, ledger work, WBR queries, or any session-start \"is this Mac ready\" question — trigger on \"preflight\", \"is the mac ready\", \"are the containers up\", \"why is psql failing\". Do NOT use to FIX anything it finds (it is read-only by design — starting containers, freeing disk, or re-authing wrangler are separate, deliberate actions), and do NOT use for device-hardware probes (sysctl/ devicectl belong to cwc-ios-engineering's device-hardware / this plugin's claude-app-artifacts).", "summary": "One read-only preflight for THIS Mac before any fleet work — seven checks, one JSON object out, each check grounded in a real incident this repo already paid for: disk free (the 2026-07-01/2026-07-03 100%-full incidents; Xcode-beta artifacts were the culprit), the subagentjobs-postgres + subagent...", "gated": 0, "source_path": "plugins/cwc-hardware-engineering/skills/mac-fleet-preflight/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_mac_fleet_preflight" }, { "id": "skl_network_capture", "name": "network-capture", "plugin": "cwc-hardware-engineering", "description": "Capture network metrics the way the Claude macOS app itself does — a copycat of its DevTools Network panel and Help > Record Net Log — into an append-only orjson JSONL hot path and an in-memory SQLite fact/dim star schema. `uv run netcap ingest-har|ingest-netlog|ingest-cdp --jsonl

` (append), `uv run netcap load --jsonl

` (drain to SQLite + print the \"N requests, X kB\" footer), `uv run netcap query --jsonl

--resource-type xhr`. Data shape = the Chrome DevTools Protocol `Network.ResourceType` enum (byte-for-byte the Network-panel filter chips and the connected chrome-devtools MCP). Use when the user asks to capture / record / trace network requests, ingest a HAR or NetLog, tap CDP traffic, or measure a page's request metrics. Trigger on \"capture network\", \"record net log\", \"HAR\", \"network requests\", \"CDP\", \"DevTools network\", \"request metrics\". Do NOT use for packet-level tracing (packet-trace) or visualizing the captured data (cwc-analytics-engineering's network-analytics reads what this writes).", "summary": "Capture network metrics the way the Claude macOS app itself does — a copycat of its DevTools Network panel and Help > Record Net Log — into an append-only orjson JSONL hot path and an in-memory SQLite fact/dim star schema.", "gated": 0, "source_path": "plugins/cwc-hardware-engineering/skills/network-capture/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_network_capture" }, { "id": "skl_agent_workforce", "name": "agent-workforce", "plugin": "cwc-human-resources", "description": "Agent-workforce (HR) playbook for the subagentjobs.com family — the \"employees\" here are typed agents and skill-based coworkers, not humans. Use when the user asks to \"hire an agent\", \"add a new agent\", \"define a new role\", \"onboard a coworker\", \"review agent performance\", \"which model does agent X run on\", \"upgrade an agent to Fable 5\", \"who is on the team\", \"org chart\", \"agent roster\", \"headcount\", \"comp review for the agents\", or anything mapping HR vocabulary onto this repo's agent fleet. Do NOT use for human people-ops (recruiting pipelines, offers, human comp, onboarding checklists for people) — this startup is a solo founder and no human HR data exists in this repo. Do NOT use for building or deploying the underlying workers (that is cwc-deploy's job), or for legal/compliance review (cwc-legal).", "summary": "Agent-workforce (HR) playbook for the subagentjobs.com family — the \"employees\" here are typed agents and skill-based coworkers, not humans.", "gated": 0, "source_path": "plugins/cwc-human-resources/skills/agent-workforce/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_agent_workforce" }, { "id": "skl_coworker_census", "name": "coworker-census", "plugin": "cwc-human-resources", "description": "The org chart of the AI workforce, generated from real files — one census row per plugins/cwc-*/ coworker: IDENTITY.json loaded STRICTLY via the canonical Pydantic mirror (scripts/coworkers/schema.py, with an honest permissive fallback), skills on disk vs plugin.json-registered, memory/current version, and gate names from gates.toml (both fleet formats). Use when the user asks \"who is on the team\", \"org chart\", \"coworker roster\", \"headcount\", \"census\", \"which coworkers have unregistered skills\", \"what gates does each coworker have\", or \"what memory version is each coworker on\". Do NOT use for hiring/review/comp mechanics (that is this plugin's agent-workforce skill), for fixing catalog drift it surfaces (cwc-product-management's skill-catalog-audit owns the diff detail; plugins/_schema/generate.py owns the regeneration), or for the .claude/agents/*.yaml generated-agent roster (agent-workforce covers those employees).", "summary": "The org chart of the AI workforce, generated from real files — one census row per plugins/cwc-*/ coworker: IDENTITY.json loaded STRICTLY via the canonical Pydantic mirror (scripts/coworkers/schema.py, with an honest permissive fallback), skills on disk vs plugin.json-registered, memory/current ve...", "gated": 0, "source_path": "plugins/cwc-human-resources/skills/coworker-census/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_coworker_census" }, { "id": "skl_apple_container_internals", "name": "apple-container-internals", "plugin": "cwc-ios-engineering", "description": "Architecture-level knowledge of Apple's `container` runtime source (vendors/apple/container) — how it actually runs a Linux container as a per-container lightweight VM: the apiserver/XPC-helper process model, the DNSServer + SocketForwarder networking path, ContainerBuild, ContainerOS, the config.toml schema, and the CLI command surface. Use when the user asks how `container` works internally, which process does what, how container networking/ DNS/port-forwarding is implemented, how to read or extend a specific subsystem under vendors/apple/container/Sources, what config.toml keys exist, or wants an exact CLI command from the vendored command-reference. Trigger on \"how does apple container work\", \"container-apiserver\", \"container networking internals\", \"DNSServer\", \"SocketForwarder\", \"vmnet\", \"container XPC\", \"config.toml\", \"container-network-vmnet\", \"container command reference\". Do NOT use for container lifecycle GATES, sandboxing policy, or eBPF/security work (apple-linux-containers owns those), nor for authoring Swift apps (swift-app-engineering).", "summary": "Architecture-level knowledge of Apple's `container` runtime source (vendors/apple/container) — how it actually runs a Linux container as a per-container lightweight VM: the apiserver/XPC-helper process model, the DNSServer + SocketForwarder networking path, ContainerBuild, ContainerOS, the config...", "gated": 0, "source_path": "plugins/cwc-ios-engineering/skills/apple-container-internals/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_apple_container_internals" }, { "id": "skl_apple_linux_containers", "name": "apple-linux-containers", "plugin": "cwc-ios-engineering", "description": "Linux systems and security engineering on Apple silicon via Apple's `container` runtime (vendors/apple/container) — OCI images run as lightweight per-container VMs on Virtualization.framework. Use when the user asks to run Linux containers or VMs on a Mac, sandbox a coworker/Claude Code session (desktop.md's bypass-permissions best practice: \"only use in sandboxed containers or VMs\"), work with OCI images outside Docker, investigate the desktop-linux Claude Desktop app inside a VM, or do OS-internals/security work in a Linux guest (eBPF, kernel instrumentation, detection-as-code, hardening). Trigger on \"apple container\", \"container system start\", \"Linux VM on Mac\", \"Virtualization.framework\", \"sandbox this session\", \"OCI on macOS\", \"eBPF\", \"kernel instrumentation\", \"seccomp\", \"detection as code\". Do NOT use for Docker Desktop operations (different runtime, different daemon) or for Cloudflare Workers sandboxes (cloud-side, not this Mac).", "summary": "Linux systems and security engineering on Apple silicon via Apple's `container` runtime (vendors/apple/container) — OCI images run as lightweight per-container VMs on Virtualization.framework.", "gated": 0, "source_path": "plugins/cwc-ios-engineering/skills/apple-linux-containers/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_apple_linux_containers" }, { "id": "skl_device_hardware", "name": "device-hardware", "plugin": "cwc-ios-engineering", "description": "Hardware introspection for Apple devices — READS the real machines and the real SDK instead of recalling. `uv run hwprobe mac` (this Mac's chip/cores/memory via sysctl), `uv run hwprobe device --udid ` (a tethered iOS device via `xcrun devicectl`), `uv run hwprobe sdk` (the version-exact nw_interface_type_t enum parsed from Network.framework's on-disk header). Every fact is provenance-tagged — probe / sdk_header / doc / unavailable — and a fact that can't be probed or cited is marked UNAVAILABLE with a reason, NEVER guessed (Apple SoC die internals are not authoritatively knowable here and are reported as such). Use when the user asks what chip/cores/memory a Mac has, what a tethered iPhone/iPad actually is, the exact Network.framework interface-type enum for the installed SDK, or wants hardware facts grounded rather than recalled. Trigger on \"what chip\", \"how many cores\", \"device info\", \"devicectl\", \"sysctl hardware\", \"nw_interface_type\", \"SDK header\", \"iPhone hardware\", \"Apple silicon cores\". Do NOT use for packet capture (packet-trace) or Swift app code (swift-app-engineering).", "summary": "Hardware introspection for Apple devices — READS the real machines and the real SDK instead of recalling.", "gated": 0, "source_path": "plugins/cwc-ios-engineering/skills/device-hardware/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_device_hardware" }, { "id": "skl_foundation_models_bridge", "name": "foundation-models-bridge", "plugin": "cwc-ios-engineering", "description": "Deep knowledge of `vendors/anthropics/ClaudeForFoundationModels` — the package that conforms Claude to Apple's FoundationModels `LanguageModel` protocol so you drive Claude through the SAME `LanguageModelSession` API (`respond(to:)`, streaming, `@Generable` guided generation, tool calling) as Apple's on-device model. THE key fact this skill exists to teach: this bridge is SERVER-SIDE (a network call to Anthropic, OS 27 / Xcode 27 / API key required) — it is NOT on-device and NOT offline; do not confuse it with Apple's own local model or with the direct ClaudeAPI Messages client. Use when the ask involves FoundationModels + Claude, `ClaudeLanguageModel`, `LanguageModelSession` with Claude, `@Generable` output from Claude, `serverTools` (web search/fetch/code execution), effort mapping, or choosing between on-device / bridge / direct-API. Trigger on \"Foundation Models\", \"FoundationModels\", \"LanguageModelSession\", \"ClaudeLanguageModel\", \"@Generable\", \"on-device model\", \"Apple Intelligence\", \"server-side language model\", \"ClaudeForFoundationModels\". Do NOT use for the direct ClaudeAPI Messages client or SwiftUI work (swift-app-engineering), or for container/VM internals (apple-container-internals).", "summary": "Deep knowledge of `vendors/anthropics/ClaudeForFoundationModels` — the package that conforms Claude to Apple's FoundationModels `LanguageModel` protocol so you drive Claude through the SAME `LanguageModelSession` API (`respond(to:)`, streaming, `@Generable` guided generation, tool calling) as App...", "gated": 0, "source_path": "plugins/cwc-ios-engineering/skills/foundation-models-bridge/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_foundation_models_bridge" }, { "id": "skl_packet_trace", "name": "packet-trace", "plugin": "cwc-ios-engineering", "description": "Trace a network packet's path from the iPhone Claude app, across home Wi-Fi, all the way to the Claude web app running in a browser on the user's Mac — and back. uv-managed typed tooling: capture-surface PLANNERS that emit real, copy-pasteable rvictl + tcpdump commands (you cannot tcpdump a stock iPhone — Apple's Remote Virtual Interface mirrors the tethered phone onto the Mac), runtime host-OS detection (never a hardcoded macOS/iOS version), and 5-tuple flow correlation that stitches separate pcaps into one end-to-end trace. Run `uv run netpath doctor` (detect host + attached iPhone UDIDs), `uv run netpath plan --udid ` (emit the capture recipes), `uv run netpath correlate --hops ` (orient the captures into flows). Trigger on \"trace the packet\", \"capture iPhone traffic\", \"rvictl\", \"tcpdump the phone\", \"where does my Claude app traffic go\", \"network path iPhone to Mac\", \"packet trace across devices\". Do NOT use for Swift app code (swift-app-engineering) or VM lifecycle gates (apple-linux-containers) — though this skill's VM_BRIDGE surface reuses the Lima/apple-container interfaces that skill owns.", "summary": "Trace a network packet's path from the iPhone Claude app, across home Wi-Fi, all the way to the Claude web app running in a browser on the user's Mac — and back.", "gated": 0, "source_path": "plugins/cwc-ios-engineering/skills/packet-trace/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_packet_trace" }, { "id": "skl_swift_app_engineering", "name": "swift-app-engineering", "plugin": "cwc-ios-engineering", "description": "iOS/Swift app engineering for this repo's Apple surfaces — the house ClaudeAPI Swift package (apps/coworkers-desktop-buddy/Sources/ClaudeAPI) and any SwiftUI/UIKit application work above it. Use when the user asks to build, extend, or debug Swift/SwiftUI/UIKit code, touch MessagesRequest/Configuration/ SSEParser in the ClaudeAPI module, optimize the mobile stack (rendering, memory, launch time, streaming), or prepare an app for TestFlight. Trigger on \"Swift\", \"SwiftUI\", \"UIKit\", \"Xcode\", \"iOS app\", \"ClaudeAPI module\", \"TestFlight\", \"app performance\". Do NOT use for: the ClaudeForFoundationModels server-side bridge or Apple's FoundationModels framework (foundation-models-bridge owns those, and they are NOT the same thing as this Messages client); shipping Cloudflare Workers (cwc-deploy); or flashing ESP32/M5Stack hardware (m5-onboard/ cardputer-buddy) — this is the Swift app-client/UI layer.", "summary": "iOS/Swift app engineering for this repo's Apple surfaces — the house ClaudeAPI Swift package (apps/coworkers-desktop-buddy/Sources/ClaudeAPI) and any SwiftUI/UIKit application work above it.", "gated": 0, "source_path": "plugins/cwc-ios-engineering/skills/swift-app-engineering/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_swift_app_engineering" }, { "id": "skl_swift_conduit", "name": "swift-conduit", "plugin": "cwc-ios-engineering", "description": "The conduit's typed Python tooling — macOS ⇄ Linux-VM Swift environments as Pydantic models, OODA loops, and SemVer metadata. Use when the user asks to probe an environment for Swift/claude-desktop (EnvironmentObserver), plan a Swift-toolchain or Claude Desktop Linux install (installers emit the verbatim desktop-linux.md apt sequence as typed InstallSteps — plans, not executions), run a build check with a typed report, record an observe→orient→decide→act iteration (uv run ooda), or read/write versions.json (MetadataStore; the Swift CLI and MCP semver_metadata tool read the same file — proven cross-language round-trip). Trigger on \"conduit\", \"Swift on Linux\", \"provision the VM\", \"install claude-desktop\", \"ooda\", \"versions manifest\", \"environment probe\". Do NOT use for authoring Swift code (swift-app-engineering) or for VM-runtime lifecycle decisions (apple-linux-containers owns those gates).", "summary": "The conduit's typed Python tooling — macOS ⇄ Linux-VM Swift environments as Pydantic models, OODA loops, and SemVer metadata.", "gated": 0, "source_path": "plugins/cwc-ios-engineering/skills/swift-conduit/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_swift_conduit" }, { "id": "skl_swift_package_inventory", "name": "swift-package-inventory", "plugin": "cwc-ios-engineering", "description": "Typed inventory of a real Swift package WITHOUT building it: runs `swift package dump-package` (manifest evaluation only — seconds, no compile), parses SwiftPM's verbose dump JSON into a compact typed summary (products, targets + their dependencies, package dependencies with version requirements), and greps MCP `Tool(name: \"...\")` registrations out of Sources/ so the MCP surface is inventoried alongside the package graph. Point it at any package dir; defaults to apps/subagentswift-mcp (the fleet's Swift MCP server). Use when the user asks what a Swift package ships, its products/targets/deps, which MCP tools a Swift server registers, or wants a package graph without waiting for a build. Trigger on \"dump-package\", \"package inventory\", \"what does this package ship\", \"which MCP tools\", \"SwiftPM graph\". Do NOT use for building/testing the package (swift-conduit's build checks own that), for the ClaudeForFoundationModels graph narrative (cwc-swift-engineering's swift-stack-trace owns that reading), or for non-SwiftPM projects.", "summary": "Typed inventory of a real Swift package WITHOUT building it: runs `swift package dump-package` (manifest evaluation only — seconds, no compile), parses SwiftPM's verbose dump JSON into a compact typed summary (products, targets + their dependencies, package dependencies with version requirements)...", "gated": 0, "source_path": "plugins/cwc-ios-engineering/skills/swift-package-inventory/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_swift_package_inventory" }, { "id": "skl_tdd_red_green", "name": "tdd-red-green", "plugin": "cwc-ios-engineering", "description": "The fleet's TDD orchestration pipeline — model-routed red/green/refactor with hard budgets and two no-silent-failure alert paths. Use when building Python (or any testable) modules coworker-style: the main-loop model (Fable 5) writes the pytest suite first (RED), a budgeted background Sonnet 5 coder agent runs the ralph-loop contract to make it pass (GREEN) then refactor, the main loop re-verifies, and a Sonnet 5 helper updates the durable plan file. Trigger on \"tdd\", \"red/green\", \"ralph\", \"write tests first\", \"dispatch a coder\", \"background implementation\", or any conduit-plan W-task marked RED/GREEN. Do NOT use for one-line fixes or docs-only changes — the harness overhead only pays for real modules with real contracts.", "summary": "The fleet's TDD orchestration pipeline — model-routed red/green/refactor with hard budgets and two no-silent-failure alert paths.", "gated": 0, "source_path": "plugins/cwc-ios-engineering/skills/tdd-red-green/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_tdd_red_green" }, { "id": "skl_compliance_playbook", "name": "compliance-playbook", "plugin": "cwc-legal", "description": "Legal/compliance playbook for the subagentjobs.com fleet. Use when the user asks \"is it OK to republish these docs\", \"review the mirrored-docs attribution\", \"check upstream ToS\", \"license audit\", \"are our LICENSE files consistent\", \"review the monetization terms\", \"merchant of record\", \"can we send this NDA/letter/notice\", \"review before we publish this policy page\", or any question about what the fleet republishes, licenses, or sends externally. Do NOT use for drafting product copy (cwc-product-management), for deploying anything (cwc-deploy), or as a substitute for qualified legal counsel — every output here is workflow assistance, not legal advice.", "summary": "Legal/compliance playbook for the subagentjobs.com fleet.", "gated": 0, "source_path": "plugins/cwc-legal/skills/compliance-playbook/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_compliance_playbook" }, { "id": "skl_license_inventory", "name": "license-inventory", "plugin": "cwc-legal", "description": "READ-ONLY typed inventory of the repo's REAL license landscape, every count computed live from the tree (never hardcoded): LICENSE files across plugins/cwc-*, \"license\" fields across workers/*/package.json (a real, fleet-wide 0-of-97 gap at the 2026-07-09 baseline — flagged honestly, not papered over), top-level vendored orgs under vendors/ (governed by their own upstream licenses), and the vendored Apache-2.0 reference-only skills inside cwc-design. Emits a JSON inventory or a text report with a findings section. Use when the user asks \"license audit\", \"are our LICENSE files consistent\", \"which workers declare a license\", \"what's vendored under what license\", \"run the license inventory\", or before any relicensing/enrollment decision that needs the current license facts. Do NOT use this to interpret license compatibility or give legal opinions — it counts and flags, and every output is workflow assistance, not legal advice (compliance-playbook owns the review discipline; qualified counsel owns conclusions).", "summary": "READ-ONLY typed inventory of the repo's REAL license landscape, every count computed live from the tree (never hardcoded): LICENSE files across plugins/cwc-*, \"license\" fields across workers/*/package.json (a real, fleet-wide 0-of-97 gap at the 2026-07-09 baseline — flagged honestly, not papered...", "gated": 0, "source_path": "plugins/cwc-legal/skills/license-inventory/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_license_inventory" }, { "id": "skl_secrets_sweep", "name": "secrets-sweep", "plugin": "cwc-legal", "description": "READ-ONLY secrets/PII exposure sweep for cwc-legal (wave W5.2) over the operator's OWN deployed subagent*.com fleet. Defensive security audit only: derives the live host list from cloudflare.toml's [dns] table (never a hardcoded guess), issues read-only HTTP GETs against each public endpoint (`/`, `/api`, `/llms.txt`), and scans each response with a conservative offline regex detector for Anthropic/Cloudflare/AWS keys, PEM private-key headers, write-secret-env-vars WITH an assigned value, contextual bearer tokens (a long token-looking string near an authorization/bearer/token/secret keyword), and non-public emails. Emits a deterministic, fully-redacted markdown report — no full secret or the operator's own write-secrets ever appear in the report itself. No writes, no exploitation, no auth bypass, no third-party targets. Use when the user asks to \"sweep the fleet for leaked secrets\", \"check if any worker leaks ANTHROPIC_API_KEY/cfat_/PII\", \"run the secrets sweep\", \"audit our own endpoints for exposure\", or references `secrets-sweep sweep`/`scan-file`. Do NOT use this for penetration testing, auth bypass, fuzzing, or auditing any host the operator does not own — those are out of scope for this skill and for cwc-legal generally.", "summary": "READ-ONLY secrets/PII exposure sweep for cwc-legal (wave W5.2) over the operator's OWN deployed subagent*.com fleet.", "gated": 0, "source_path": "plugins/cwc-legal/skills/secrets-sweep/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_secrets_sweep" }, { "id": "skl_startup_formation", "name": "startup-formation", "plugin": "cwc-legal", "description": "Pre-formation legal groundwork for this AI startup: US business licensing/entity research path, trademark and branding audit across the ~50 subagent*.com domains, copyright posture for the fleet's original-vs-mirrored content, and terms review for startup-program applications (Cloudflare Workers Launchpad). Use when the user asks \"business license\", \"form the company\", \"LLC or C-corp\", \"trademark\", \"can I use the Claude/Cloudflare name\", \"copyright my content\", \"review the launchpad terms\", \"legal groundwork\", or before ANY program application is submitted. Do NOT use for reviewing what the fleet republishes day-to-day (sibling compliance-playbook skill) — and NOTHING here is legal advice; see the hard rule.", "summary": "Pre-formation legal groundwork for this AI startup: US business licensing/entity research path, trademark and branding audit across the ~50 subagent*.com domains, copyright posture for the fleet's original-vs-mirrored content, and terms review for startup-program applications (Cloudflare Workers...", "gated": 0, "source_path": "plugins/cwc-legal/skills/startup-formation/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_startup_formation" }, { "id": "skl_agent_traffic_flywheel", "name": "agent-traffic-flywheel", "plugin": "cwc-marketing", "description": "The growth model for agentic traffic to the ~50-domain subagent*.com fleet: a six-stage self-reinforcing flywheel (discoverability → cheap visit → cross-link traversal → durable value → return traffic → citation → discovery again), with ten concrete levers and locally-measurable metrics. Use when the user asks \"grow agent traffic\", \"the flywheel\", \"spin the flywheel\", \"flywheel metrics\", \"why aren't agents visiting\", \"what should we ship next for traffic\", or wants the traffic-growth backlog prioritized. Monetization of that traffic (Monetization Gateway, x402, pay-per-crawl) is cwc-finance's monetization-gateway skill — this skill grows the traffic that makes monetization worth anything. Do NOT use for one-off content copywriting (sibling content-and-discoverability skill) or for recording analytics snapshots (cwc-analytics-engineering).", "summary": "The growth model for agentic traffic to the ~50-domain subagent*.com fleet: a six-stage self-reinforcing flywheel (discoverability → cheap visit → cross-link traversal → durable value → return traffic → citation → discovery again), with ten concrete levers and locally-measurable metrics.", "gated": 0, "source_path": "plugins/cwc-marketing/skills/agent-traffic-flywheel/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_agent_traffic_flywheel" }, { "id": "skl_content_and_discoverability", "name": "content-and-discoverability", "plugin": "cwc-marketing", "description": "Marketing playbook for the subagent*.com fleet — keep the fleet's agent-readiness surfaces (per-site /llms.txt, robots.txt with Content-Signal, sitemap.xml on subagentsubdomains.com, /.well-known/api-catalog on subagentapi.com) accurate as marketing assets for agent traffic; draft on-brand copy for new sites by reading subagentbrands.com's recorded guidelines; and announce new primitives in subagentnews.com's changelog voice. Use when the user asks \"audit our llms.txt\", \"is the sitemap current\", \"draft copy for the new site\", \"announce the new primitive\", \"check discoverability\", \"marketing pass on the fleet\", \"is this site discoverable by agents\", or \"does this copy match our voice\". Do NOT use for recording or editing brand_guidelines rows — that is cwc-product-management's brand-voice skill's job; this skill only READS guidelines. Do NOT use for deploying anything — publishing routes through cwc-deploy's playbook and is gated.", "summary": "Marketing playbook for the subagent*.com fleet — keep the fleet's agent-readiness surfaces (per-site /llms.txt, robots.txt with Content-Signal, sitemap.xml on subagentsubdomains.com, /.well-known/api-catalog on subagentapi.com) accurate as marketing assets for agent traffic; draft on-brand copy f...", "gated": 0, "source_path": "plugins/cwc-marketing/skills/content-and-discoverability/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_content_and_discoverability" }, { "id": "skl_creative_production", "name": "creative-production", "plugin": "cwc-marketing", "description": "Agency-grade creative production for the subagent*.com brand: palette-locked art direction, written briefs before pixels, batch variations, provenance-tagged AI imagery via the Higgsfield MCP connector (when reachable), and publish gates. Use when the operator asks for deck visuals, og-images, campaign imagery, video loops, or \"make it look professional\". Direction is the deliverable; generators are interchangeable.", "summary": "Agency-grade creative production for the subagent*.com brand: palette-locked art direction, written briefs before pixels, batch variations, provenance-tagged AI imagery via the Higgsfield MCP connector (when reachable), and publish gates.", "gated": 0, "source_path": "plugins/cwc-marketing/skills/creative-production/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_creative_production" }, { "id": "skl_flywheel_weekly", "name": "flywheel-weekly", "plugin": "cwc-marketing", "description": "The weekly agent-traffic flywheel pass for the subagent*.com fleet (36+ live sites optimized for AI-agent traffic): baseline metrics via scripts/flywheel-metrics.sh, a deterministic 6-site agent-surface drift audit (audit.py), public traffic/trend context from the cwc-marketing toolkit profile's cloudflare-radar server, and 1-2 brand-grounded content drafts into reports/marketing/. Use when the user asks for the \"weekly flywheel\", a \"marketing pass\", a \"content audit\", an \"agent traffic report\", or when a scheduled weekly marketing run fires. Do NOT use for the flywheel model/levers themselves (sibling agent-traffic-flywheel skill), for one-off copy or announcement drafting outside the weekly cadence (sibling content-and-discoverability skill), or for deploying anything — publishing stays behind gates.toml's content_publish gate and routes through cwc-deploy.", "summary": "The weekly agent-traffic flywheel pass for the subagent*.com fleet (36+ live sites optimized for AI-agent traffic): baseline metrics via scripts/flywheel-metrics.sh, a deterministic 6-site agent-surface drift audit (audit.py), public traffic/trend context from the cwc-marketing toolkit profile's...", "gated": 0, "source_path": "plugins/cwc-marketing/skills/flywheel-weekly/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_flywheel_weekly" }, { "id": "skl_launch_notes", "name": "launch-notes", "plugin": "cwc-marketing", "description": "Render grounded launch announcements from the REAL fleet deploy ledger (fact_fleet_deploy in the local Postgres warehouse) instead of marketing fluff: one markdown note per deploy wave — what shipped, counts by pattern, and a verification statement — with the hard rule enforced in code that a domain may only appear if its ledger row is status='verified'. Never announce unverified work. Use when the user asks to \"announce the expansion\", \"write launch notes\", \"draft the wave announcement\", \"what shipped this wave\", or after a fleet deploy wave lands and its verification pass completes. Do NOT use for publishing — this skill only RENDERS text; landing it on agentchangelogs.com or agentbloggers.com is a deploy behind gates.toml's content_publish gate, routed through cwc-deploy. Do NOT use for one-off copy or weekly-cadence content (sibling content-and-discoverability and flywheel-weekly skills own those).", "summary": "Render grounded launch announcements from the REAL fleet deploy ledger (fact_fleet_deploy in the local Postgres warehouse) instead of marketing fluff: one markdown note per deploy wave — what shipped, counts by pattern, and a verification statement — with the hard rule enforced in code that a dom...", "gated": 0, "source_path": "plugins/cwc-marketing/skills/launch-notes/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_launch_notes" }, { "id": "skl_dependency_drift_audit", "name": "dependency-drift-audit", "plugin": "cwc-operations", "description": "Fleet-wide dependency-drift audit for the workers/ Cloudflare Worker fleet: the committed dependency graph (scripts/build-graph.py — kit imports + runtime-fetch edges), the per-worker kit-drift lock gate (scripts/worker-build.sh — .build-lock.json vs current _shared/kit hashes), and the dead-runtime-fetch-target check. Use when the user says \"drift audit\", \"stale workers\", \"which workers need redeploy\", \"kit changed\", \"I edited shell.ts/css.ts/d1crud.ts/http.ts/ wellknown.ts/mcp.ts\", \"who imports the kit\", \"who fetches this site\", \"blast radius\", \"redeploy queue\", \"check runtime fetch targets\", \"refresh the build graph\", or before marking any fleet task complete after a kit edit. Also fire as the weekly-review drift step (sibling fleet-operations skill, step 5). Do NOT use for executing the redeploys themselves (route to cwc-deploy), for design-token/CSS audits (`scripts/design-system-audit.sh` — sibling fleet-audit skill), or for the 36 tsk_resync_ nav-drift tasks (content drift, not build drift).", "summary": "Fleet-wide dependency-drift audit for the workers/ Cloudflare Worker fleet: the committed dependency graph (scripts/build-graph.py — kit imports + runtime-fetch edges), the per-worker kit-drift lock gate (scripts/worker-build.sh — .build-lock.json vs current _shared/kit hashes), and the dead-runt...", "gated": 0, "source_path": "plugins/cwc-operations/skills/dependency-drift-audit/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_dependency_drift_audit" }, { "id": "skl_fleet_audit", "name": "fleet-audit", "plugin": "cwc-operations", "description": "Fleet-audit toolkit for the subagent*.com sites. Use when the user asks to \"audit the fleet\", \"run the design-system audit\", \"check site health\", \"health check the primitives\", \"sweep zone ids\", \"which zones are unrouted\", \"check for drift\", \"triage the resync tasks\", or after any fleet-wide change that needs verification. Do NOT use for fixing what the audit finds (route fixes to cwc-engineering to build and cwc-deploy to ship) and do NOT use for the routing/review loop itself — that is the sibling fleet-operations skill.", "summary": "Fleet-audit toolkit for the subagent*.com sites.", "gated": 0, "source_path": "plugins/cwc-operations/skills/fleet-audit/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_fleet_audit" }, { "id": "skl_fleet_daily_digest", "name": "fleet-daily-digest", "plugin": "cwc-operations", "description": "The single daily ops read for the subagent*.com fleet (~50 Cloudflare workers) — one pass, one file, five fixed sections. Use when the user asks for the \"ops digest\", \"daily fleet check\", \"morning digest\", \"fleet status\", \"what's the fleet look like today\", or a scheduled daily run fires. Do NOT use for deep drift audits or resync triage (sibling fleet-audit skill), for incident response mid-breakage (sibling incident-runbook), or for the routing/review loop itself (sibling fleet-operations) — this skill only produces the one daily read and names ONE action; executing that action routes to the owning plugin.", "summary": "The single daily ops read for the subagent*.com fleet (~50 Cloudflare workers) — one pass, one file, five fixed sections.", "gated": 0, "source_path": "plugins/cwc-operations/skills/fleet-daily-digest/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_fleet_daily_digest" }, { "id": "skl_fleet_operations", "name": "fleet-operations", "plugin": "cwc-operations", "description": "Operations playbook for running the subagent*.com fleet day to day. Use when the user asks \"operations status\", \"run the operating loop\", \"route this to a coworker\", \"which coworker owns this\", \"weekly review\", \"operate the fleet\", \"how many sites do we have\", \"what's open across the fleet\", or wants operational work decomposed into contracts/tasks/outcomes on the family's own primitives. Do NOT use for executing domain work itself — that belongs to the owning domain plugin (cwc-design, cwc-data, cwc-deploy, cwc-analytics-engineering, cwc-product-management, cwc-engineering, cwc-finance) — and do NOT use for running the audit scripts themselves; that is the sibling fleet-audit skill.", "summary": "Operations playbook for running the subagent*.com fleet day to day.", "gated": 0, "source_path": "plugins/cwc-operations/skills/fleet-operations/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_fleet_operations" }, { "id": "skl_incident_runbook", "name": "incident-runbook", "plugin": "cwc-operations", "description": "The fleet's incident runbook + change discipline, adapted from Anthropic's operations plugin (runbook, change-request, risk-assessment, status-report) and grounded in this fleet's real incident history. Use when something breaks mid-wave, when a deploy misbehaves, before system-state changes, or on \"incident\", \"runbook\", \"what went wrong\", \"postmortem\".", "summary": "The fleet's incident runbook + change discipline, adapted from Anthropic's operations plugin (runbook, change-request, risk-assessment, status-report) and grounded in this fleet's real incident history.", "gated": 0, "source_path": "plugins/cwc-operations/skills/incident-runbook/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_incident_runbook" }, { "id": "skl_registry_consistency_audit", "name": "registry-consistency-audit", "plugin": "cwc-operations", "description": "Cross-check one domain's state (or the whole fleet's) across the FOUR registries that must agree: the fleet manifest (plugins/cwc-deploy/skills/fleet-expansion/fleet-manifest.toml — [sites.*]/[redirects.*]/[denylist] incl. never_use), the Postgres deploy ledger (fact_fleet_deploy via psql -U wbr), the live dns_zones API (subagentsubdomains.com/api/dns-zones), and the repo-root cloudflare.toml [dns] section — and emit a per-domain consistency matrix with a single verdict (CONSISTENT-TOMBSTONE / CONSISTENT-LIVE / CONSISTENT-UNSERVED / INCONSISTENT / UNKNOWN). Born from the real 2026-07-09 subagentcloudflare.com teardown, where one operator directive had to land in all four places at once. Use when the user asks \"is consistent across the registries\", \"did the teardown/launch land everywhere\", \"audit the registries\", \"registry drift\", after any fleet-expansion registration wave, or after tombstoning/denylisting a domain. Do NOT use for nav/SITES-array drift on the sites themselves (sibling dependency-drift-audit / fleet-audit own that), for fixing what it finds (route to cwc-deploy / cwc-engineering), or for design-system gates (scripts/design-system-audit.sh).", "summary": "Cross-check one domain's state (or the whole fleet's) across the FOUR registries that must agree: the fleet manifest (plugins/cwc-deploy/skills/fleet-expansion/fleet-manifest.toml — [sites.*]/[redirects.*]/[denylist] incl.", "gated": 0, "source_path": "plugins/cwc-operations/skills/registry-consistency-audit/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_registry_consistency_audit" }, { "id": "skl_brand_voice", "name": "brand-voice", "plugin": "cwc-product-management", "description": "Brand-voice playbook for the subagentjobs.com family — read subagentbrands.com's existing brand_guidelines, and when asked to add a new one, ground it in a real quoted example pulled from an actual sibling-site file (never invented), then record it durably after operator confirmation. Also covers tailoring an on-brand message to its actual audience (a coworker, the operator, an external reader) without drifting off-voice in the process. Use when the user asks \"discover brand voice\", \"what is our brand voice\", \"check brand guidelines\", \"audit brand voice\", \"add a brand guideline\", \"is this on-brand\", \"brand voice for subagentjobs\", \"does this match our voice\", \"brand voice audit\", \"write this update for \", or pastes a piece of copy and asks whether it fits the family's tone. Do NOT use this to search Notion, Confluence, Google Drive, Box, Microsoft 365, Figma, Gong, Granola, or Slack — none of those platforms are connected in this environment; see \"What's different in this adaptation\" before assuming enterprise-platform discovery applies here.", "summary": "Brand-voice playbook for the subagentjobs.com family — read subagentbrands.com's existing brand_guidelines, and when asked to add a new one, ground it in a real quoted example pulled from an actual sibling-site file (never invented), then record it durably after operator confirmation.", "gated": 1, "source_path": "plugins/cwc-product-management/skills/brand-voice/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_brand_voice" }, { "id": "skl_citation_curation", "name": "citation-curation", "plugin": "cwc-product-management", "description": "Record verbatim, programmatically-verified quotes from the mirrored docs corpora as durable, doubly-typed citation rows in this plugin's citations/*.citations.jsonl — each row carries BOTH the Anthropic Citations API shape (1:1 with subagentcitations.com's D1 columns, char_location) AND a real CSL-JSON item validated against the vendored citation-style-language schema (vendors/citation-style-language/schema, mirrored in this skill's references/csl-data.json). Preloaded by the three sonnet-5 curator agents (cowork-/claude-tag-/office-agents-citation- curator); also usable directly. Use when the user says \"cite the doc for X\", \"store that quote\", \"curate citations\", \"verify the citation store\", \"publish citations\". All row I/O goes through scripts/cite.py (extract/add/verify/publish-body) — the script refuses any quote that is not an exact substring of the source file. Do NOT use for cataloging/organizing docs (docs-librarian skill) or for paraphrases — verbatim quotes only.", "summary": "Record verbatim, programmatically-verified quotes from the mirrored docs corpora as durable, doubly-typed citation rows in this plugin's citations/*.citations.jsonl — each row carries BOTH the Anthropic Citations API shape (1:1 with subagentcitations.com's D1 columns, char_location) AND a real CS...", "gated": 0, "source_path": "plugins/cwc-product-management/skills/citation-curation/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_citation_curation" }, { "id": "skl_citations_indexeddb", "name": "citations-indexeddb", "plugin": "cwc-product-management", "description": "Python client + ingester for the IndexedDB-pattern store pinned in schemas/INDEXEDDB-COWORKERS.md — the durable, queryable home of the operator-chosen 11 gh_orgs datasets (scripts/chrome-output) and every citations/*.citations.jsonl row, at ~/Library/Application Support/Coworkers/IndexedDB (the Claude.app-copycat location; jsonl records, deliberately NOT LevelDB). Use when the user wants to ingest/refresh the store (\"ingest the datasets\", \"refresh the grounding store\"), run grounding lookups (\"resolve this citation\", \"which repos ground cit_X\", \"query gh_repositories by language/org/stars\"), or check store health (\"verify the store\", exit 2 on any spec-invariant failure). All I/O goes through `uv run citations-indexeddb ingest|query|resolve|verify`. Swift parity: the same contract's client is CoworkersModel/GitHubOrgs.swift; tests/test_swift_parity.py reads it at test time. Do NOT use for authoring/verifying verbatim quotes (citation-curation's cite.py owns that), and never write into Claude.app's own IndexedDB directory or modify scripts/chrome-output.", "summary": "Python client + ingester for the IndexedDB-pattern store pinned in schemas/INDEXEDDB-COWORKERS.md — the durable, queryable home of the operator-chosen 11 gh_orgs datasets (scripts/chrome-output) and every citations/*.citations.jsonl row, at ~/Library/Application Support/Coworkers/IndexedDB (the C...", "gated": 0, "source_path": "plugins/cwc-product-management/skills/citations-indexeddb/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_citations_indexeddb" }, { "id": "skl_dependency_contracts", "name": "dependency-contracts", "plugin": "cwc-product-management", "description": "The PRODUCT view of cross-worker dependencies in the subagent*.com fleet: which runtime_fetch edges in workers/build-graph.json are intentional product surface (backed by a durable contract row on subagentcontracts.com) vs incidental coupling that merely works today. Use when the user asks \"who depends on this route\", \"can we break/rename this API\", \"add a cross-site fetch\", \"dependency contracts\", \"uncontracted edges\", \"is this dependency contracted\", wants the add-a-dependency or deprecate-a-route policy, or wants the uncontracted-edge backlog worked. Do NOT use for kit-import build staleness or rebuild sequencing (`worker-build.sh --stale` is cwc-engineering's discipline — this skill only cites it), for implementing the fetch code itself (cwc-engineering), or for shipping the deploy (cwc-deploy).", "summary": "The PRODUCT view of cross-worker dependencies in the subagent*.com fleet: which runtime_fetch edges in workers/build-graph.json are intentional product surface (backed by a durable contract row on subagentcontracts.com) vs incidental coupling that merely works today.", "gated": 0, "source_path": "plugins/cwc-product-management/skills/dependency-contracts/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_dependency_contracts" }, { "id": "skl_docs_librarian", "name": "docs-librarian", "plugin": "cwc-product-management", "description": "Maintain the plugin's library/ catalogs — JSON indexes mapping every doc in the three mirrored corpora (cowork+plugins+skills, claude-tag+connectors, office-agents under docs/docs/claude.com/docs/) to the cwc-product-management skills, src/coworkers Swift surfaces, subagent*.com primitives, and memory/vN directives it grounds. Preloaded by the three haiku librarian agents (cowork-/claude-tag-/office-agents-librarian); also usable directly. Use when the user says \"organize the docs\", \"refresh the catalog\", \"librarian pass\", \"which doc grounds X\", \"catalog coverage\". All catalog I/O goes through scripts/catalog.py (files/status/upsert/ validate) — deterministic, stdlib-only, schema-enforcing. Do NOT use for pulling verbatim citation rows (citation-curation skill) or for editing the docs themselves — corpora are read-only.", "summary": "Maintain the plugin's library/ catalogs — JSON indexes mapping every doc in the three mirrored corpora (cowork+plugins+skills, claude-tag+connectors, office-agents under docs/docs/claude.com/docs/) to the cwc-product-management skills, src/coworkers Swift surfaces, subagent*.com primitives, and m...", "gated": 0, "source_path": "plugins/cwc-product-management/skills/docs-librarian/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_docs_librarian" }, { "id": "skl_feature_spec", "name": "feature-spec", "plugin": "cwc-product-management", "description": "Write a feature spec for subagentjobs work BEFORE it becomes cards on the Coworkers board (src/coworkers) or an edge in dependency-contracts — problem statement, goals/non-goals, and P0/P1/P2 requirements that map directly onto real Card(content, priority, owner) rows, not a generic PRD template. Use when the user asks to \"spec this feature\", \"write a PRD\", \"scope this idea\", \"what should this feature actually do\", \"break this into P0/P1/P2\", or hands PM a vague ask that needs turning into board-ready cards before sprint-board can assign it a lane. Do NOT use for roadmap sequencing across multiple features (product-strategy), for an already-scoped cross-worker API promise (dependency-contracts), or for assigning already-written cards to lanes (sprint-board) — this skill's job ends where a spec becomes a set of Card rows ready to place.", "summary": "Write a feature spec for subagentjobs work BEFORE it becomes cards on the Coworkers board (src/coworkers) or an edge in dependency-contracts — problem statement, goals/non-goals, and P0/P1/P2 requirements that map directly onto real Card(content, priority, owner) rows, not a generic PRD template.", "gated": 0, "source_path": "plugins/cwc-product-management/skills/feature-spec/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_feature_spec" }, { "id": "skl_fivetran_rollout_decomposition", "name": "fivetran-rollout-decomposition", "plugin": "cwc-product-management", "description": "Decompose the coworkers/ Fivetran multi-connector data-storage project (Slack, cloudflare-analytics, github, gmail connectors + connector_registry/connector_runs/source_tables in coworkers/data-infra) into atomic tasks with subtasks, and own the one green-light approval gate before any task is handed to cwc-engineering for implementation. Use when the user asks to \"decompose the fivetran rollout\", \"break down the connector project\", \"what tasks are next\", \"what's the next atomic task for fivetran\", \"plan the connector rollout\", \"greenlight the next connector task\", or asks for the task/subtask breakdown of wiring Slack or cloudflare-analytics to real data. Do NOT use for narrating why this project exists or keeping the registry-counts/ doc-provenance story current (cwc-analytics-engineering's data-vision-runbook skill owns the vision narrative) and do NOT use for writing connector/schema/API code yourself (cwc-engineering's engineering-playbook owns implementation) — this skill's job ends the moment a task is approved and hands off.", "summary": "Decompose the coworkers/ Fivetran multi-connector data-storage project (Slack, cloudflare-analytics, github, gmail connectors + connector_registry/connector_runs/source_tables in coworkers/data-infra) into atomic tasks with subtasks, and own the one green-light approval gate before any task is ha...", "gated": 0, "source_path": "plugins/cwc-product-management/skills/fivetran-rollout-decomposition/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_fivetran_rollout_decomposition" }, { "id": "skl_product_strategy", "name": "product-strategy", "plugin": "cwc-product-management", "description": "Product-strategy discipline for the subagent*.com business: positioning (user ≠ buyer), one named north-star metric with guardrails, and a now/next/later roadmap where every item carries its gate owner (finance spend gate, legal attorney gate, PM north-star check). Use when shaping strategy artifacts — investor decks, roadmaps, positioning docs — or when reviewing finance/legal outputs for product coherence. Trigger on \"product strategy\", \"positioning\", \"north star\", \"roadmap\", \"now/next/later\", or when a deck/plan mixes audience, metric, and sequencing questions.", "summary": "Product-strategy discipline for the subagent*.com business: positioning (user ≠ buyer), one named north-star metric with guardrails, and a now/next/later roadmap where every item carries its gate owner (finance spend gate, legal attorney gate, PM north-star check).", "gated": 0, "source_path": "plugins/cwc-product-management/skills/product-strategy/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_product_strategy" }, { "id": "skl_session_documentation", "name": "session-documentation", "plugin": "cwc-product-management", "description": "The PRODUCT-MANAGEMENT workflow that turns a finished work session into a durable, browsable record in apps/coworkers-console/sessions/ — deciding a session is worth documenting, driving cwc-design's session-record-html scaffold, filling it with verified narrative anchored to the session's own real trace (traceId/spanId from src/coworkers' Telemetry.swift, when the session ran through the Coworkers board), and placing the session's plan/tracker artifacts alongside it under the session identifier. Use when the user says \"record this session\", \"document what we did\", \"put the artifacts in coworkers-console\", \"session record for \", or after a multi-task plan completes and its evidence should outlive the chat. Do NOT reimplement the HTML template or the scaffold script — those belong to cwc-design (skills/session-record-html); this skill orchestrates them and owns file placement + the anti-fabrication gate. Do NOT use for the live coworkers-console dashboard (index.html + serve.py), which is engineering's.", "summary": "The PRODUCT-MANAGEMENT workflow that turns a finished work session into a durable, browsable record in apps/coworkers-console/sessions/ — deciding a session is worth documenting, driving cwc-design's session-record-html scaffold, filling it with verified narrative anchored to the session's own re...", "gated": 0, "source_path": "plugins/cwc-product-management/skills/session-documentation/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_session_documentation" }, { "id": "skl_skill_catalog_audit", "name": "skill-catalog-audit", "plugin": "cwc-product-management", "description": "Catalog-integrity audit across every cwc-* coworker plugin: diffs (a) skills/ dirs on disk vs (b) the plugin.json \"skills\" array vs (c) IDENTITY.json skills[] names, reporting unregistered dirs (16 SKILL.md-bearing ones existed on 2026-07-09 — the motivating incident), dead plugin.json entries pointing at missing dirs, and IDENTITY staleness that needs a plugins/_schema/generate.py rerun. Use when the user asks \"are all skills registered\", \"audit the skill catalog\", \"is plugin.json in sync with skills/\", \"is IDENTITY.json stale\", \"which skills are orphaned\", or before/after a batch that adds or removes skill directories. Do NOT use to fix the drift it reports (edit each plugin.json by hand; IDENTITY/BRAND are machine-generated — rerun plugins/_schema/generate.py + validate.py, never hand-edit them), to validate IDENTITY shape/content (that is validate.py + scripts/coworkers/schema.py), or for the fleet-wide roster overview (cwc-human-resources' coworker-census).", "summary": "Catalog-integrity audit across every cwc-* coworker plugin: diffs (a) skills/ dirs on disk vs (b) the plugin.json \"skills\" array vs (c) IDENTITY.json skills[] names, reporting unregistered dirs (16 SKILL.md-bearing ones existed on 2026-07-09 — the motivating incident), dead plugin.json entries po...", "gated": 0, "source_path": "plugins/cwc-product-management/skills/skill-catalog-audit/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_skill_catalog_audit" }, { "id": "skl_sprint_board", "name": "sprint-board", "plugin": "cwc-product-management", "description": "Run the weekly product-management sprint-planning cadence against the REAL Coworkers board (src/coworkers — a live, tested SwiftUI app, not a spec) — jointly with cwc-data-engineering. Sets Sprint.goal (Sun→Sat UTC, computed by SprintWeek.bounds), assigns backlog Cards to coworker lanes via Board.lane(for:), and reviews the weekly session-telemetry the board itself appends to telemetry.jsonl. Grounded in the actual Cowork VM session/data model (docs/cowork/3p/data-storage.md) that CoworkersStore copycats. Use when the user asks to plan the sprint, set weekly strategy, assign tasks to coworkers, review the weekly board/telemetry metrics, or groom the backlog. Trigger on \"sprint planning\", \"weekly strategy\", \"assign tasks\", \"board review\", \"backlog grooming\", \"Coworkers board\". Do NOT use for the data model itself (that's cwc-data-engineering's session-telemetry-erd) or the SwiftUI view code (cwc-swift-engineering) — this skill operates the board, it doesn't build it.", "summary": "Run the weekly product-management sprint-planning cadence against the REAL Coworkers board (src/coworkers — a live, tested SwiftUI app, not a spec) — jointly with cwc-data-engineering.", "gated": 0, "source_path": "plugins/cwc-product-management/skills/sprint-board/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_sprint_board" }, { "id": "skl_wbr_deep_dive", "name": "wbr-deep-dive", "plugin": "cwc-product-management", "description": "The PM's Monday-to-Wednesday deep-dive cadence on top of the WBR warehouse: Monday 10:00 PST cwc-data-engineering delivers reports/wbr/wbr-.xlsx (their SLA); Mon/Tue the PM reads it and leads with the 3-5 largest |WoW| exceptions, consulting any coworker's domain (skills/, memory/, reports// digests, live APIs) with every consultation recorded per finding; Wednesday 10:00 PST ships reports/wbr/.md in the prior week's voice, with ranked recommendations that each name a dim_metric metric_id, a gate owner, and the disposition ledger of last week's recommendations. Every number is quoted from a workbook cell or an rpt_wbr row — the model computes nothing. Trigger on \"wbr deep dive\", \"monday deep dive\", \"wednesday report\", \"weekly report\". Extends cwc-data's weekly-business-review discipline — never replaces it.", "summary": "The PM's Monday-to-Wednesday deep-dive cadence on top of the WBR warehouse: Monday 10:00 PST cwc-data-engineering delivers reports/wbr/wbr-.xlsx (their SLA); Mon/Tue the PM reads it and leads with the 3-5 largest |WoW| exceptions, consulting any coworker's domain (skills/, memory/, re...", "gated": 0, "source_path": "plugins/cwc-product-management/skills/wbr-deep-dive/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_wbr_deep_dive" }, { "id": "skl_agent_gtm", "name": "agent-gtm", "plugin": "cwc-sales", "description": "Go-to-market discipline for selling content and APIs to AI agents: the scored account book (50 zones ranked for monetization readiness), the Gateway enrollment pipeline, and the weekly forecast line into the WBR. Use when the operator says \"GTM\", \"sales\", \"monetize\", \"score the book\", \"pipeline\", or when monetization strategy needs an account-level view instead of a fleet-level one. The buyer is the agent's operator; the user is the agent — sales owns the buyer motion.", "summary": "Go-to-market discipline for selling content and APIs to AI agents: the scored account book (50 zones ranked for monetization readiness), the Gateway enrollment pipeline, and the weekly forecast line into the WBR.", "gated": 0, "source_path": "plugins/cwc-sales/skills/agent-gtm/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_agent_gtm" }, { "id": "skl_pipeline_outreach", "name": "pipeline-outreach", "plugin": "cwc-sales", "description": "The daily sales pipeline pass for the subagent*.com fleet — snapshot the cwc-sales ledger from subagenttasks.com/subagentcontracts.com (pipeline.py, deterministic and read-only), qualify each open row against the fleet's actual sellable inventory, and batch outreach DRAFTS into reports/sales/outreach-drafts-YYYYMMDD.md — never sending anything; the outreach_send gate in gates.toml governs every draft, always. Use when the user says \"run the pipeline\", \"qualify leads\", \"draft outreach\", \"sales pipeline pass\", or wants the day's batch of drafts done in one pass. One pass per day max. Do NOT use for logging a single new prospect or one-off qualification discipline — that's pipeline-playbook. Do NOT use for pricing/monetization mechanics — cwc-finance's monetization-gateway skill owns the Cloudflare Monetization Gateway pilot.", "summary": "The daily sales pipeline pass for the subagent*.com fleet — snapshot the cwc-sales ledger from subagenttasks.com/subagentcontracts.com (pipeline.py, deterministic and read-only), qualify each open row against the fleet's actual sellable inventory, and batch outreach DRAFTS into reports/sales/outr...", "gated": 0, "source_path": "plugins/cwc-sales/skills/pipeline-outreach/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_pipeline_outreach" }, { "id": "skl_pipeline_playbook", "name": "pipeline-playbook", "plugin": "cwc-sales", "description": "Sales playbook for the subagent*.com fleet — maintain a durable prospect/opportunity ledger as real tasks/contracts rows on subagenttasks.com and subagentcontracts.com, qualify inbound interest in the fleet's primitive APIs, and draft outreach that is always operator-gated before sending. Use when the user asks \"track this prospect\", \"what's in the pipeline\", \"qualify this inbound\", \"draft outreach to X\", \"sales status\", \"who's interested in the APIs\", \"log this opportunity\", or \"follow up with the lead\". Do NOT use for pricing/monetization mechanics — the Cloudflare Monetization Gateway pilot is owned by cwc-finance's monetization-gateway skill. Do NOT use expecting a CRM, email integration, or contact connector — none exist in this environment, and this skill says so instead of simulating them.", "summary": "Sales playbook for the subagent*.com fleet — maintain a durable prospect/opportunity ledger as real tasks/contracts rows on subagenttasks.com and subagentcontracts.com, qualify inbound interest in the fleet's primitive APIs, and draft outreach that is always operator-gated before sending.", "gated": 0, "source_path": "plugins/cwc-sales/skills/pipeline-playbook/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_pipeline_playbook" }, { "id": "skl_prospect_agent_readiness", "name": "prospect-agent-readiness", "plugin": "cwc-sales", "description": "Scored agent-readiness audit of ANY external domain — sales-qualification tooling in agent-gtm's scored-account-book discipline: six read-only checks over public well-known discovery surfaces (llms.txt, robots.txt Content-Signal, Web Bot Auth key directory, MCP server card, x402/AP2 experiments, homepage rel=alternate link), weighted per this skill's own declared rubric into a 0-100 score with a grade. Use when qualifying a prospect or buyer account (\"how agent-ready is example.com\", \"score this domain\", \"audit their agent surfaces\", \"readiness check before the pitch\"), or to rank a batch of prospect domains for outreach priority. Do NOT use for the fleet's own sites (scripts/flywheel-metrics.sh and cwc-marketing own fleet-side measurement), for contacting the prospect (outreach stays behind gates.toml's outreach_send gate), or for anything beyond polite read-only GETs — this skill never POSTs, never crawls, never retries.", "summary": "Scored agent-readiness audit of ANY external domain — sales-qualification tooling in agent-gtm's scored-account-book discipline: six read-only checks over public well-known discovery surfaces (llms.txt, robots.txt Content-Signal, Web Bot Auth key directory, MCP server card, x402/AP2 experiments,...", "gated": 0, "source_path": "plugins/cwc-sales/skills/prospect-agent-readiness/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_prospect_agent_readiness" }, { "id": "skl_bio_tools_catalog", "name": "bio-tools-catalog", "plugin": "cwc-science", "description": "Typed catalog of Claude Science's 24 bundled bio-tools MCP servers (pubmed, biomart, chembl, clinical_trials, genomes, variants, and 18 more — 466 real tools total, live-verified 2026-07-04) plus an on-demand invoker script. NOT a standing MCP registration — nothing here loads into every agent turn. Use when a task needs real biomedical/genomic/chemistry data (a PubMed search, a gene cross-reference, a ChEMBL compound lookup, a clinical trial search, etc.) and Claude Science is installed on this Mac. Read tools.json to find the right package/tool, then invoke it once via call_tool.py. Requires Claude Science (Claude Bioscience) to be installed — check with plugins/cwc-engineering/skills/platform-doctor first if unsure.", "summary": "Typed catalog of Claude Science's 24 bundled bio-tools MCP servers (pubmed, biomart, chembl, clinical_trials, genomes, variants, and 18 more — 466 real tools total, live-verified 2026-07-04) plus an on-demand invoker script.", "gated": 0, "source_path": "plugins/cwc-science/skills/bio-tools-catalog/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_bio_tools_catalog" }, { "id": "skl_determinism_harness", "name": "determinism-harness", "plugin": "cwc-science", "description": "Double-run determinism verification for any pipeline, generator, or seeded computation: runs a shell command TWICE in fresh temp working dirs, diffs the declared output files (byte-identical by default, or a normalized mode that first strips caller-declared volatile regex patterns like timestamps/run-ids), and returns a typed verdict {deterministic, first_divergent_file, diff_excerpt} plus a CLI with meaningful exit codes (0/1/2). Generalizes this repo's two real double-run precedents: the corpus pipeline's run-twice byte-identical reproducibility check and scripts/build-graph.py's byte-identical-on-unchanged-tree artifact discipline. Use when the user asks \"is this pipeline deterministic\", \"verify reproducibility\", \"run it twice and diff\", \"check the generator doesn't churn\", or before committing any generated artifact whose diff should mean something. Do NOT use for flaky-TEST hunting (run the test runner's own repeat tooling), for cross-machine reproducibility (this harness only certifies same-machine double-run identity — an honest, narrower claim), or on commands with side effects outside their working directory (each run gets a fresh temp cwd; external state is not isolated).", "summary": "Double-run determinism verification for any pipeline, generator, or seeded computation: runs a shell command TWICE in fresh temp working dirs, diffs the declared output files (byte-identical by default, or a normalized mode that first strips caller-declared volatile regex patterns like timestamps...", "gated": 0, "source_path": "plugins/cwc-science/skills/determinism-harness/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_determinism_harness" }, { "id": "skl_reproducible_analysis", "name": "reproducible-analysis", "plugin": "cwc-science", "description": "Reproducible scientific analysis grounded in Claude Science (Anthropic's AI workbench for scientists): every artifact carries the exact code, environment, and history that produced it; a reviewer pass checks citations, numbers, and figure-vs-code agreement; working pipelines are saved as reusable skills and version-controlled in this repo. Use on \"analyze\", \"reproduce\", \"review this figure/citation\", \"save this pipeline as a skill\", or any Claude Science work. Compute that hits SSH/HPC/Modal is operator-gated.", "summary": "Reproducible scientific analysis grounded in Claude Science (Anthropic's AI workbench for scientists): every artifact carries the exact code, environment, and history that produced it; a reviewer pass checks citations, numbers, and figure-vs-code agreement; working pipelines are saved as reusable...", "gated": 0, "source_path": "plugins/cwc-science/skills/reproducible-analysis/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_reproducible_analysis" }, { "id": "skl_agent_schema_swift_mirror", "name": "agent-schema-swift-mirror", "plugin": "cwc-swift-engineering", "description": "The Swift (5th-language) mirror of the fleet's typed agent schema — real `enum AgentModel: String, Codable, CaseIterable` and `enum AgentTool` whose rawValues exactly mirror crates/schema/src/agent.rs (serde), scripts/agents/schema.ts (Zod), and scripts/agents/schema.py (Pydantic) — plus an executable cross-language drift check (compile with swiftc, print rawValues as JSON, diff against schema.py). Use when the user asks for the agent schema in Swift, to check the Swift mirror for drift, to add a model or tool to the shared agent enums, or after any edit to agent.rs/schema.ts/ schema.py. Trigger on \"AgentModel in Swift\", \"swift mirror\", \"schema drift\", \"5th language\", \"cross-language check\". Do NOT use for the Messages-API wire shape (swift-stack-trace owns CodingKeys/ClaudeAPI), for regenerating .claude/agents/*.yaml (that's `cargo run -p agent-gen`), or for editing the sibling mirrors themselves without also updating this one.", "summary": "The Swift (5th-language) mirror of the fleet's typed agent schema — real `enum AgentModel: String, Codable, CaseIterable` and `enum AgentTool` whose rawValues exactly mirror crates/schema/src/agent.rs (serde), scripts/agents/schema.ts (Zod), and scripts/agents/schema.py (Pydantic) — plus an execu...", "gated": 0, "source_path": "plugins/cwc-swift-engineering/skills/agent-schema-swift-mirror/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_agent_schema_swift_mirror" }, { "id": "skl_swift_stack_trace", "name": "swift-stack-trace", "plugin": "cwc-swift-engineering", "description": "Trace the real Swift stack the Claude apps and this repo already ship — the language, the SwiftPM package graph, and the frameworks — before writing new Swift. Copycat-first groundwork: read the actual sources, don't invent. Covers the house ClaudeAPI Messages client (Configuration/MessagesRequest/SSEParser), the ClaudeForFoundationModels bridge package layout, the official modelcontextprotocol/swift-sdk dependency, and the one hard rule — the wire shape is shared across four languages (Swift CodingKeys mirror serde/Pydantic/ Zod). Use when the user asks to build/extend Swift code, map a SwiftPM graph, touch CodingKeys/Codable, work with Swift concurrency or SSE, or understand the real Swift surface. Trigger on \"Swift\", \"SwiftPM\", \"swift package\", \"CodingKeys\", \"Codable\", \"async/await Swift\", \"SSE parser\". Do NOT use for device hardware (cwc-hardware-engineering), the Xcode toolchain (cwc-xcode-engineering), or Cloudflare deploys.", "summary": "Trace the real Swift stack the Claude apps and this repo already ship — the language, the SwiftPM package graph, and the frameworks — before writing new Swift.", "gated": 0, "source_path": "plugins/cwc-swift-engineering/skills/swift-stack-trace/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_swift_stack_trace" }, { "id": "skl_xcode_footprint_audit", "name": "xcode-footprint-audit", "plugin": "cwc-xcode-engineering", "description": "Audit the real Xcode-toolchain disk footprint on this Mac as typed JSON — what IS installed (xcode-select path, Swift toolchain, /Applications/ Xcode*.app glob, simctl availability) and what disk the support artifacts actually hold (DerivedData, iOS DeviceSupport, CoreSimulator), each check {id, present, detail, disk_bytes?} with honest absence, never assuming anything exists. The \"check before blaming\" tool for disk-full incidents: this Mac's recurring 100%-full episodes were Xcode-beta + its artifacts, while Docker.raw and Time Machine snapshots were false leads. Use when disk is filling up, before deleting anything Xcode-shaped, when asked \"is Xcode installed\", \"what's eating my disk\", \"how big is DerivedData\", or before/ after installing or removing an Xcode. Trigger on \"footprint\", \"disk audit\", \"DerivedData size\", \"DeviceSupport\", \"simulator disk\". Do NOT use for toolchain version/SDK probing before builds (xcode-toolchain-trace owns that) or for actually deleting artifacts — this skill is read-only; deletion is the operator's call.", "summary": "Audit the real Xcode-toolchain disk footprint on this Mac as typed JSON — what IS installed (xcode-select path, Swift toolchain, /Applications/ Xcode*.app glob, simctl availability) and what disk the support artifacts actually hold (DerivedData, iOS DeviceSupport, CoreSimulator), each check {id,...", "gated": 0, "source_path": "plugins/cwc-xcode-engineering/skills/xcode-footprint-audit/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_xcode_footprint_audit" }, { "id": "skl_xcode_toolchain_trace", "name": "xcode-toolchain-trace", "plugin": "cwc-xcode-engineering", "description": "Trace the real installed Xcode toolchain before acting — the SDKs, the xcrun command surface (build/test/simulate/sign/distribute), and the native Xcode Source Editor Extension ecosystem. Copycat-first: probe `xcodebuild -version`, `-showsdks`, and `xcrun --find `; never recall a version. Grounded in the vendored theswiftdev/awesome-xcode-extensions list. Use when the user asks to build/test/sign/distribute an Xcode project, which SDKs/tools are installed, how to run the simulator, notarize or upload a build, or which Xcode editor extensions exist for formatting/CodingKeys/lint. Trigger on \"Xcode\", \"xcodebuild\", \"xcrun\", \"SDK\", \"simulator\", \"simctl\", \"notarize\", \"TestFlight\", \"provisioning\", \"Xcode extension\". Do NOT use for Swift language/package work (cwc-swift-engineering) or device-hardware introspection (cwc-hardware-engineering).", "summary": "Trace the real installed Xcode toolchain before acting — the SDKs, the xcrun command surface (build/test/simulate/sign/distribute), and the native Xcode Source Editor Extension ecosystem.", "gated": 0, "source_path": "plugins/cwc-xcode-engineering/skills/xcode-toolchain-trace/SKILL.md", "created_at": "2026-07-10 18:13:13", "cite_as": "https://subagentskills.com/api/skills/skl_xcode_toolchain_trace" } ] }