Claude Skills by yabafre

Use when user says "research idea", "aped analyze", or invokes aped-analyze. Runs alongside aped-context — both can apply on hybrid projects (new feature in existing system).

Use when user says "audit architecture", "find shallow modules", "deepen modules", "deletion test", "depth analysis", "leverage analysis", "scan for refactor opportunities", "improve codebase architecture", "pass-through wrappers", "audit module depth", "aped arch audit", or invokes aped-arch-audit. Surfaces deepening candidates in an existing codebase — produces a report and HALTs; never auto-refactors.

Use when user says "create architecture", "technical architecture", "solution design", "system design", "design the architecture", or invokes aped-arch. Runs between PRD and Epics.

Use when user says "brainstorm", "ideate", "generate ideas", "divergent thinking", or invokes aped-brainstorm. Horizontal — invokable at any phase.

Use when user says "checkpoint", "review changes", "walk me through this", "show me the diff", "walk the diff", "what just happened", "summarize the changes", or invokes aped-checkpoint. Not for sprint progress dashboards — see aped-status for that.

Use when user says "update CLAUDE.md", "sync claude rules", "aped claude", or invokes aped-claude. Merges — never overwrites user customizations.

Brownfield entry-point — use when user says "document codebase", "project context", "existing project", "existing codebase", "legacy project", "onboarding to a codebase", "aped context", or invokes aped-context. The default first step for any brownfield project before aped-analyze; runs alongside aped-analyze on hybrid projects (new feature in a legacy system).

Use when user says "correct course", "change scope", "pivot", "aped correct", or invokes aped-course.

Use when user says "debug", "diagnose", "troubleshoot", "why is X failing", "find the root cause", "feedback loop", "hypothesise", "aped debug", or invokes aped-debug. Also invoked from aped-dev on persistent test red (≥3 failed attempts) and from aped-review on findings that need root-cause investigation.

Use when user says "design twice", "alternative design", "explore options", "what else could we do", "second opinion on architecture", or invokes aped-design-twice. Generates two competing designs before committing to one. Prevents premature lock-in.

Use when user says "start dev", "implement story", "aped dev", "TDD", "TDD cycle", "red green refactor", "failing test first", "generate unit tests", or invokes aped-dev. Not for hotfixes or single-file isolated changes — see aped-quick for those. Not for E2E tests — see aped-qa for that.

Use when user says "discuss epic", "discuss epic N", "lock decisions for epic", "epic implementation decisions", "SPIDR for epic", "what should we decide for this epic", or invokes aped-discuss-epic. Locks per-epic implementation decisions BEFORE aped-story spawns story files — fills the gap between aped-arch (system-level) and aped-story (per-story). Not for system-wide design (use aped-arch) or per-story plan (use aped-story).

Use when user says "elicit", "critique this", "stress-test", "deeper review", "pre-mortem", "red team", "socratic", "tree of thoughts", "first principles", or invokes aped-elicit. Horizontal — invokable at any phase.

Use when user says "create epics", "break into stories", "aped epics", or invokes aped-epics. Does NOT create story files — that''s aped-story.

Use when user says "from ticket", "pickup ticket", "pickup from Linear", "pickup from Jira", "work on this ticket", "start the GitHub issue", "ingest ticket", references an external ticket ID, "aped from-ticket", or invokes aped-from-ticket.

Use when user says "build glossary", "update glossary", "domain dictionary", "shared language", "sharpen language", "canonicalize terms", "glossary check", or invokes aped-glossary. Builds and maintains a project-wide domain glossary at docs/aped/glossary.md so every APED skill, persona, and downstream artefact uses the same vocabulary. Acts as the project-wide CONTEXT.md analog. Not for divergent ideation (use aped-brainstorm), not for stress-testing decisions (use aped-grill).

Use when user says "grill me", "align on this", "interview me about", "get on the same page", "pin down the spec", "stress test the plan", or invokes aped-grill. Relentless one-question-at-a-time alignment. Stops when (a) no new question for two turns, (b) token budget exceeds 25k, (c) user says stop. Emits grill-summary.md for downstream skills (aped-prd, aped-arch). Not for divergent ideation — see aped-brainstorm. Not for critique-method elicitation — see aped-elicit.

Use when user says "iterate", "rework", "after merge", "post-ship", "what now after merge", "we shipped X, now we need Y", "address the gap", or invokes aped-iterate. Routes a post-ship delta to the right downstream skill — does not implement the change itself.

Use when user says "lead", "check approvals", "check sprint check-ins", "aped lead", or invokes aped-lead. Runs from the main project, not a worktree.

Use when user says "create PRD", "generate PRD", "draft requirements", "product requirement", "write the prd", "aped prd", or invokes aped-prd. Headless mode available via --headless.

Use when user says "pre-mortem", "what could go wrong", "risk analysis", "failure modes", "anticipate problems", or invokes aped-pre-mortem. Structured pre-mortem before implementation starts. Not for post-incident (see aped-retro).

Use when user says "prfaq", "working backwards", "press release first", "stress-test the concept", or invokes aped-prfaq. Optional upstream tooling — can run before aped-analyze.

Use when user says "purge aped", "clean docs", "tidy aped", "audit docs/aped", "what is in docs/aped", "regenerate aped index", "find orphan docs", "aped doc hygiene", or invokes aped-purge. Walks docs/aped/ (or the configured output_path), classifies every entry as canonical / unknown / archived, regenerates the INDEX.md entry point, and surfaces orphan files (docs Claude wrote outside the APED canon) with a per-file triage menu. Read-only by default — moves and deletes only happen on user c...

Use when user says "generate tests" (E2E or integration), "E2E tests", "integration tests", "aped qa", or invokes aped-qa. Not for unit tests written during TDD RED — those are owned by aped-dev.

Use when user says "quick fix", "quick feature", "hotfix", "small change", "just fix this", "aped quick", or invokes aped-quick. Bypasses the full A→P→E→D→R pipeline — use only for isolated fixes.

Use when receiving code review feedback or when aped-dev hands back review comments to address. Invoked from aped-dev after aped-review reports issues, or standalone when user pastes external review feedback (PR comments, GitHub thread, senior engineer Slack message). Triggers on phrases like "address review", "address PR comments", "fix the review feedback", "the reviewer said", "received PR comments".

Use when user says "retro", "retrospective", "epic done — what did we learn", "aped retro", or invokes aped-retro. Runs after the last story of an epic ships.

Use when user says "review code", "run review", "aped review", or invokes aped-review.

Use when user says "ship", "sprint pr", "pre-push", "aped ship", "open the umbrella PR", "end of sprint", "composite review", or invokes aped-ship. Only runs from the main project on the base branch — never from inside a worktree. HALTs before push.

Use when user says "parallel sprint", "dispatch stories", "aped sprint", or invokes aped-sprint. Only runs inside the main project, not inside a worktree. Creates worktrees only — story-ready and state flips are owned by aped-story.

Use when user says "sprint status", "show progress", "aped status", or invokes aped-status. Not for walking through a diff or summarising recent changes — see aped-checkpoint for that.

Use when user says "create story", "prepare next story", "aped story", or invokes aped-story.

Use when user says "triage this issue", "classify this bug", "is this in scope", "out of scope", "should we fix this now", "prioritize", "triage", or invokes aped-triage. Issue triage state machine: intake → classify → route → close/defer. Maintains .out-of-scope/ knowledge base for deferred items. Emits triage-decision.md for downstream skills.

Use when user says "design UX", "UX spec", "aped ux", or invokes aped-ux. Runs between PRD and Epics.

Use when user says "write a skill", "create a skill", "new APED skill", "extend APED with my own skill", "make a custom skill", "add a custom slash command", "scaffold a skill", or invokes aped-write-skill. Meta-skill: walks the user through writing a Claude Code skill that follows APED conventions. Not for routing-time invocation discipline — see aped-claude for that.

Use when user says "zoom out", "step back", "take a step back", "broader view", "wider context", "lost the thread", "are we still solving the right problem", or invokes aped-zoom-out. Horizontal — invokable at any phase. Not for sprint progress dashboards (see aped-status) or for diff walks (see aped-checkpoint).