Agent
Root Cause Analyzer. Consumes an Evidence Bundle (from the investigator) and determines the actual root cause and mechanism — without re-scanning the whole repo. Returns a structured Root Cause with fix options (no code). Use for "explain the root cause", "why does this happen", after evidence is gathered.
Agent
Validation Agent (adversarial verifier). Verifies a change does what it should — derives its OWN edge cases/holdout, exercises them, checks regressions and cross-service contract breaks. Consumes a change (diff/Design Spec) as UNTRUSTED input and returns a Validation Report with a clear verdict + unforgeable evidence.…
Skill Claude CodeCodex
Navigate the codebase cheaply by reading the context store first, not re-scanning the repo. Use at the start of any investigation, feature, or review to locate the relevant service/file fast and with minimal tokens. Triggers on "where is", "how does X work", "understand the code", "find the service for".
Skill Claude CodeCodex
Commit workflow — before committing, sync the knowledge layer (context store, per-service CLAUDE.md, and the agents/skills routing README) to the change, then draft a proper commit message and push. Use for "commit", "commit and push", "ship this", "save my work". Always stops for explicit approval before pushing.
Skill Claude CodeCodex
Maintain and compress the context store so analysis is done once and reused, never repeated. Use to record a confirmed root cause, an architecture decision, or a service/contract change; or to compress sprawling notes back into the terse store. Triggers on "remember this", "update the context", "record decision", "we…
Skill Claude CodeCodex
Validate an agent hand-off (Evidence Bundle, Root Cause, Design Spec, Implementation Report, Validation Report, Knowledge Entry) against its data contract before consuming it. Use at any pipeline seam to enforce LOCAL-reject ("escalate, don't guess") and the unforgeable-evidence rule (a PASS must carry test counts)…
Skill Claude CodeCodex
Trace a request/data flow end-to-end across services, queues, databases, object stores, and APIs to find every component a change or bug touches. Use when a problem spans boundaries or before changing a shared contract. Triggers on "trace the flow", "what depends on", "blast radius", "end to end path", "who…
Skill Claude CodeCodex
Design a minimal-impact implementation for a new feature or fix before any code is written — files to touch, contracts, migrations, risks, and a test plan. Use for "implement X", "add feature", "how should we build this". Reuses the context store; records the decision as an ADR.
Skill Claude CodeCodex
Turn a symptom into a confirmed root cause using evidence, not guesses — across any service, DB, queue, API, or infra. Use for "find the bug", "why is this failing", "0 records / stuck / DLQ / timeout / wrong data". Checks known findings first; escalates when evidence is missing instead of assuming.
Skill Claude CodeCodex
Decide what to test and write the tests for a change — regression guards, edge cases, cross-service contracts — and scaffold the test project in the repo's conventions. Use after a fix or feature, or to assess coverage. Triggers on "write tests", "test coverage", "regression test", "validate change", "review test…
Skill Claude CodeCodex
Model any system as data nodes (schema, origin, authority) and transforms tagged pure|effect, governed by validation rules and an authority that resolves conflicts. The shared frame the other df- skills build on. Triggers on "data contract", "data flow", "validation rule", "transform", "pure or effect", "authority"…
Skill Claude CodeCodex
Dispatch sub-agents under Promise Theory: state the promise plus the exact unforgeable evidence required, withhold holdout cases, verify the evidence not the self-report. USE EVERY TIME you spawn a sub-agent. Triggers on "dispatch sub-agent", "spawn agent", "parallel agents", "delegate", "fan out".
Skill Claude CodeCodex
Dark Factory Infrastructure stage: produce the Deployment and Infrastructure Spec — DTAP placement, where data lives, trust boundaries mapped to enforcement mechanisms, secrets, and a verified observability surface. Triggers on "infrastructure", "deployment", "trust boundary", "where data lives", "deploy across…
Skill Claude CodeCodex
Dark Factory Observability stage: make the consumable surface (dashboards and queryable traces keyed to scenarios, verified rendering live data) a named deliverable — a /metrics endpoint nobody can see is not observability. Triggers on "observability", "dashboards", "Grafana", "correlationId trace", "telemetry…
Skill Claude CodeCodex
Dark Factory Product Owner stage: define Vision, Requirements as data contracts plus validation rules, and Test Scenarios — the semantics every later stage derives from. Triggers on "product owner", "requirements", "acceptance criteria", "test scenarios", "data contract", "what should we build".
Skill Claude CodeCodex
Run the Dark Factory QA stage — execute the validation rules as tests, capture unforgeable evidence per scenario (by correlationId), run the holdout, and return a Works? verdict. QA is the OBSERVER: a self-report is not an assessment. Use when testing a deployed system, mapping PO scenarios to test cases, verifying…
Skill Claude CodeCodex
Dark Factory Solution Architect stage: turn PO semantics into a Data Model, Data Flow (transform graph, pure|effect, idempotency, compensation) and Service Map, assigning each validation rule an enforcement locus. Triggers on "architecture", "data model", "data flow", "service map", "schema", "Avro contract"…
Skill Claude CodeCodex
Dark Factory Developer stage: build a service test-first (RED-GREEN-REFACTOR) where the test list is the PO validation rules and scenarios, not invented. Go and Python examples. Triggers on "TDD", "write tests", "test-first", "implement the service", "RED GREEN REFACTOR".
Skill Claude CodeCodex
Drive and verify a Clerk-protected web UI end-to-end with a backend-minted session, no human login: mint the session, emit Playwright storageState, click through capturing per-scenario evidence and a verdict. Triggers on 'verify the UI', 'click through the app', 'test the front end', 'Clerk Playwright', 'authed UI…
Skill Claude CodeCodex
Parse a video (mp4/mov/webm) into a transcript + timestamped frames so Claude can consume it, then extract context — specifically Dark Factory requirements — from it. Use when given a recording, demo, app walkthrough, screen recording, or "video" to summarize or turn into requirements/specs; when asked to "watch"…
Skill Claude CodeCodex
Evaluate how easy it is for new engineers to work in a codebase. Use when assessing code quality, onboarding friction, or technical debt. Triggers on "code quality", "maintainability", "developer experience", "DX review", "onboarding", "cognitive load", "code smell", "refactoring", "tech debt".
Skill Claude CodeCodex
Audit frontend code for security, UX regressions, performance, and accessibility issues. Use when reviewing React, Next.js, Vue, Svelte, or any frontend code changes. Triggers on "audit frontend", "review UI", "check accessibility", "a11y audit", "frontend security", "UX review", "performance audit", "review…
Skill Claude CodeCodex
Automatic session-handoff across Claude Code context compaction. Hooks keep a live handoff fresh, guarantee a deterministic floor at the compaction boundary, and re-inject it after compaction/clear so the agent resumes with intent intact — no manual /handoff. Use when setting up auto-handoff, surviving auto-compact…
Skill Claude CodeCodex
Design and architect microservices following Chris Richardson's Microservices Patterns. Use when designing services, APIs, data flows, or evaluating architecture decisions. Guides full workflow from feature intake to production-ready design with observability, async patterns, security, testing, and deployment…