Skill Claude Code
Multi-perspective codebase survey that produces beads and mermaid diagrams from code structure. Each agent perspective (prod, staging, dev, feature, PM) examines the same codebase at a different zoom level, creating tagged beads and a disagreement matrix.
Skill Claude Code needs its repo
Wire validation so local == CI by construction: every check is a Taskfile target; CI and hooks invoke it rather than reimplementing it. Use when adding a validation gate, when local/CI disagree, or when a hook duplicates a command.
Skill Claude Code
Hunt vacuous / low-fidelity tests — the ones that pass whether or not the code is wrong in the way they claim to catch. Probes four shapes: a double that dodges the real thing, an assertion too loose to constrain behavior, inputs that never reach the claim, and coverage that is claimed but not guaranteed. Uses…
Skill Claude Code needs its repo
Decompose feature work into reviewable, shippable PR-sized units before coding starts. Use for a ticket, a feature description, or before a multi-file change. Feeds writing-plans — each unit becomes a plan.
Skill Claude Code
Mine the user's Claude Code chat corpus for self-optimization patterns — repeated corrections, tool thrash, dropped threads — into a ranked, PII-scrubbed report. Different from self-audit (PR code review). Dispatch as a sub-agent to keep context clean.
Skill Claude Code
Before opening a PR, enumerate every behavior it claims — stated, callers- unaffected, invariants-preserved — and tag each with the strongest evidence actually held: RAN, READ, or ASSUMED with a closed-set reason. Ship the ASSUMED residue in the PR description; log harness blockers durably. After review, reconcile…
Skill Claude Code
Use when asked whether a repo's architecture is still right — a counterfactual "how would this be designed today?" question, a suspicion of over/under-engineering or monolith-vs-split, recurring bug families that smell structural, or before committing to a major refactor. Read-only, evidence-first, multi-pass; "the…
Skill Claude Code
Use when one interface/trait has multiple implementations that are assumed equivalent (storage backends, providers, transports), when several public surfaces (CLI/API/webhook/worker) expose the same verb, when bugs cluster as "works on X but not Y", or to scope a guardrail-parity fix before estimating it. Read-only…
Agent Claude Code
Use this agent to review a design doc, RFC, ADR, or technical proposal for the DECISION it asks you to ratify — not the prose. It applies the design-doc-review discipline: frame the decision + its reversibility, run seven lenses (problem legitimacy, alternatives, reversibility/blast-radius, invariants/failure-modes…
Skill Claude Code
Review a design doc for the decision it asks you to ratify — not the prose. Catches the expensive-to-reverse mistakes while they are still words: a problem stated as a solution, alternatives that were never really considered, a one-way door taken without noticing it is one, invariants that hold only on the happy path…
Skill Claude Code
Deterministic end-to-end implementation orchestrator. Use when building net-new code or a fix that should follow the full pipeline — decompose → research → TDD loop → self-review gate → structural review. Thin: it SEQUENCES existing skills (problem-decomposer/work-scope, feature-impl…
Skill Claude Code
Turn one dev branch into small, individually-reviewable stacked PRs that land on a feature branch (one final merge to main), using jujutsu (jj) colocated with git plus keyless commit signing. Use after pr-buckets decides the slice boundaries, or whenever a large or multi-PR ticket wants one branch you follow linearly…
Skill Claude Code
Check whether the current human-agent work episode is mechanically safe to close, park incomplete work behind a verified durable checkpoint and receipt, or inspect a previously parked episode before a fail-closed resume gate.
Skill Claude Code
Group a body of in-progress work (a diff, a branch, a set of changes) into small, individually-reviewable PR buckets sized to a LoC threshold — one concern per bucket, in dependency order. The input to jj-stack. Use when one dev branch has grown past a review-friendly size and needs slicing into small PRs. Distinct…
Skill Claude Code
Capture a note, playbook, peer, or project-context update into the user's personal work HUD — a local markdown tree at $HUDROOT. Triggers on "jot this down", "remember this for ", "add a playbook for X", "note this peer/colleague", "update the context for ", or "lint the hud". The consolidation rule: this is the only…
Skill Claude Code
Post a GitHub review comment as an applyable "suggested change" — a suggestion block the author commits with one click — and, first, decide whether a finding should be a suggestion at all. gh has no native support for inline comments or suggestions, so this drives the REST reviews API directly. Use when a finding is a…
Skill Claude Code
Triage a list of candidate findings — from a PR review, a self-audit, an adversarial agent, a counterfactual or parity audit — and decide which ones deserve to be raised. Two bars, both required: TRUE (the code really behaves that way, mechanism verified at a named line) and MATTERS (practically actionable against…
Skill Claude CodeCodex
Use when creating or updating a pull request description or body — opening a PR, or when a PR body is a raw template, a bot/scanner/generated dump, or a wall of headings a reviewer must read through.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: