skills

Use when the plan is approved, user says "create beads", "beads for...", or plan documents exist.

Use when beads are created (/beads completed), user says "review beads", "check beads", "bead review", or before starting /execute.

Use when design is approved, user says "write the plan", "plan this", "create plan for...", or a design document exists.

Use when a plan is complete (/plan finished), user says "review the plan", "check the plan", or before running /beads on a new plan.

Use when beads exist and are ready for implementation. Triggers on "execute", "start implementation", "run beads", or when ready beads exist in the tracker.

Use after /execute completes, user says "review execution", "verify beads", or before creating a PR for executed work.

Use when starting a business feature that needs requirements, when user says "write requirements", "create PRD", "define user stories", or after brainstorm/discovery approval.

Use before /technical-design, after /prd, when user says "review prd", "check requirements", or when PRD quality is uncertain.

Use when documents need quality convergence, user says "autoresearch", "converge", "fix all issues", or after a review reveals multiple findings.

Use after /execute, before /review, when user says "benchmark", "performance", "check speed", or before shipping performance-sensitive changes.

Use when starting any new feature, refactoring, or exploring approaches. Triggers on "brainstorm", "let's explore", "how should we approach", or when the user has a problem but no clear direction yet.

Use when learnings emerge during development — pattern discoveries, gotchas, architecture decisions, or process improvements. Triggers on "compound", "capture learning", after review cycles, after /review approval, or when insight is fresh.

Use when something is broken, tests fail unexpectedly, or behavior doesn't match expectations. Triggers on "this isn't working", "there's a bug", "why is this happening", "debug this", or "diagnose". Do NOT use for new features, performance optimization, or code quality improvements.

Use after brainstorm for complex features, when user says "discover requirements", "deep dive", "requirements elicitation", or when brainstorm scope classifier recommends COMPREHENSIVE.

Use when starting a new project, when user says "init", "initialize project", "set up docs", "set up workflow", or when onboarding a project to the skill pipeline. Idempotent — safe to run on existing projects.

Use after /execute, before /review, when user says "QA", "test the app", "browser test", "check the UI", or for any feature with UI changes.

Use when starting a new feature that needs competitive analysis, technical landscape review, or prior art investigation. Triggers on "research this", "what's out there for", "investigate options for", or before any major new initiative where the problem space is unfamiliar.

Use before /plan, after /technical-design, when user says "review design", "check the design", or when design quality is uncertain.

Use when implementation is complete (/execute finished), user says "review", "code review", or "check the code", before creating a PR, or after significant changes.

Use after /review, before /ship, when user says "security audit", "security review", "check for vulnerabilities", or before shipping auth, payment, or data features. READ-ONLY — never modifies code.

Use when review is complete, user says "ship", "create PR", "ready to merge", or after /review approval.

Use when user says "design the system", "create technical spec", "API design", "architecture", or after PRD approval. Also use for technical improvements after brainstorm.

Generate and update documentation using templates for any Diataxis type (tutorial, howto, reference, explanation), AI files (CLAUDE.md, AGENTS.md, llms.txt), architecture docs (C4 context/container, system overview, Mermaid diagrams), and project files (README, CONTRIBUTING, glossary). Template-driven with idempotent updates that preserve custom sections. Use when creating or updating documentation. Triggers on "write docs for", "create a tutorial", "document the API", "generate README", "create AGENTS.md", "write architecture overview".

Record architectural decisions using MADR templates with full lifecycle support (Proposed, Accepted, Deprecated, Superseded). Supports full, minimal, and bare template variants. Use when making or recording architectural decisions, documenting technology choices, or capturing decision rationale. Triggers on "record this decision", "create an ADR", "document why we chose", "capture rationale".

Analyze documentation health for existing codebases. Detects structure, applies Diataxis gap analysis, checks AI-friendliness (CLAUDE.md, AGENTS.md, llms.txt), and scans for staleness. Use when joining a project, assessing documentation quality, or before a major documentation effort. Triggers on "review my docs", "what docs are missing", "audit documentation", "check doc health".