befund

Reads the full stage_graph.json from befund-stage-map and finds god-modules, circular dependencies, and layering violations, verifying each against actual code. Use when the user asks to "find architecture deficiencies", "find god-objects or god-modules", "detect dependency cycles", "check for layering violations", or as part of befund-autopilot's CHECK phase.

Runs befund's full workflow (CHECK → PLAN → approval gate → FIX+VALIDATE) with andon-loop, halting on unproven wires or unmet blockers. Use when the user asks to "run the auto-pilot", "check, plan, fix and validate this repo", or wants the full befund value stream.

Audits git remote topology and CI/CD configuration for redundancy, mirror risk, doc-vs-config drift, and commit-signing inconsistencies, masking every credential to a short preview. Use when the user asks to "audit our git remotes", "check CI setup", "find redundant mirrors", "verify docs about CI/CD are accurate", or as part of befund-autopilot's CHECK phase.

Judges idioms against the actual language version declared in the repo's manifest, never a fixed list, and categorizes each finding as modernization or smell. Use when the user asks to "find modernization opportunities", "check for deprecated idioms", "find code smells", or as part of befund-autopilot's CHECK phase.

Computes a relative complexity/attention-priority index per stage using the fixed formula 2.94 x (KSLOC)^1.10, and lists unmeasured stages plainly rather than inventing numbers. Use when the user asks "which module needs attention first", "score complexity per stage", "what's our tech debt hotspot", or as part of befund-autopilot's CHECK phase.

Extracts falsifiable claims about current code state from CLAUDE.md, README.md, DECISIONS.md, ARCHITECTURE.md, and ADR files, and verifies each against the cited code. Use when the user asks to "check documentation accuracy", "find doc drift", "verify our docs match the code", or as part of befund-autopilot's CHECK phase. Excludes CI/CD-specific doc claims, handled by befund-ci-topology, and structural contract/docstring drift -- type signatures, docstring parameter/return declarations, or API/OpenAPI/GraphQL schemas versus their call sites -- handled by zeugnis-contract-drift.

Mines executable calculations, validations, and state transitions into Given/When/Then rules, looping to convergence and requiring a two-judge panel to confirm any P0 rule. Use when the user asks to "extract business logic", "document the domain rules hidden in this code", "turn tribal knowledge into a spec", or as part of befund-autopilot's CHECK phase. Not for code style, naming, layering, dependency-direction or other architecture conventions — use lehre-codify for those. This skill extracts what the code decides, not how it is written.

Applies only eligible modernization-category findings from code_idiom_summary.json, gated behind idiom_fix.mode="fix", one remediator dispatch per (file, kind) cluster, then hands off to andon-verify without self-verifying. Use when the user explicitly asks to "apply the modernization findings", "fix the idiom findings", or "auto-fix what code-idiom found". Not for design-handbook conformance findings from a cupertino-handbook-check pass -- use cupertino-handbook-fix for those. If no code_idiom_summary.json exists, this is not the audit the user means.

Extracts discrete, checkable rules from .claude/house-rules.md (or CLAUDE.md as a best-effort fallback) and dispatches finders to verify violations against the actual code, capping dispatch at lint_max_rules. Reports violations only -- for mechanical fixes see befund-idiom-fix. Use when the user asks to "check our conventions", "check this code against our standards", "does the code follow the conventions we wrote down", "audit against house rules", "verify code follows CLAUDE.md", or as part of befund-autopilot's CHECK phase. Not for auditing work against a cupertino domain handbook under .cupertino/ -- use cupertino:cupertino-handbook-check for that.

Grades each repo in a directory of git repositories Red/Amber/Green/Gray by worst-signal-wins, and requires an explicit portfolio directory when cwd is itself a git repo. Use when the user asks to "sweep multiple repos", "assess our whole portfolio", "grade all our projects", or names a parent directory containing several git repositories.

Verifies language detection, tool availability, smoke-parseability, house-rules presence, git/CI presence, and doc presence, then assigns a Ready/Ready-with-gaps/Not-ready verdict per downstream skill. Use when the user asks to "check if this repo is ready for befund", "run preflight", "can befund analyze this codebase", or before any other befund-* skill runs for the first time in a repo; also invoked at the start of befund-autopilot's CHECK phase.

Extracts the real import/use graph per detected language, clusters files into stages by shallowest package boundary (never by manifest directory), and writes the full stage graph other befund skills depend on. Use when the user asks to "map the repo's architecture", "show me the real module boundaries", "map stages and wires", or as the first step of befund-autopilot's CHECK phase.

Reports which befund artifacts exist, whether they're stale relative to the latest commit, and never fabricates data for a skill that has not run. Use when the user asks "where does befund stand", "what's stale in our analysis", "what should we run next", or "befund status".

Synthesizes stage-map, arch-health, and every other domain summary into a phased, ranked, read-only transformation plan. Use after befund's finding skills have run, when the user asks to "synthesize the findings into a plan", "write the modernization brief", "what should we fix and in what order", or as the PLAN step of befund-autopilot.

Applies exactly one human-authorized phase of MODERNIZATION_BRIEF.md to the codebase. Use when the user explicitly asks to "execute phase N from the modernization brief", "apply this transform phase", or "run the merge/split transform-execute proposed". It applies exactly one human-authorized phase from MODERNIZATION_BRIEF.md, gated behind transform.mode="execute", a clean tree, and every Open Question resolved -- then hands off to andon-verify without self-verifying. Not for applying a set of approved fixes across a repo, or for proving each fix as it lands -- that is andon-loop; this skill requires a numbered phase in MODERNIZATION_BRIEF.md and refuses without one.

Statically audits JSX/TSX, Vue/Svelte, HTML, and CSS/SCSS source for accessibility, semantic-markup, and design-token problems -- never running or rendering the app. Use when the user asks to "audit UI accessibility", "check our components for a11y issues", "find hardcoded design values", "review this UI before we ship", or as part of befund-autopilot's CHECK phase. Not for designing an interface that does not exist yet, and not a design critique -- for build-time design of a new screen or surface use cupertino-council, and for the end-to-end design pipeline that surface belongs to use cupertino-review.