Land one architecture decision atomically: full entry to architecture-archive.json, hard-capped lean entry to architecture.json. Handles new decisions, supersession, amendment. Triggers: 'record this decision', 'log an ADR', 'we decided', 'supersede ADR', 'that replaces the earlier decision'.
Maintain a code-derived 'current belief of the system architecture' map — key applications, datastores, contracts, and integration points. Triggers: 'map architecture', 'architecture map', 'what does the system look like', 'refresh architecture map'.
Update CLAUDE.md with learnings, preferences, or project context. Use when user says "update claude.md", "add to claude.md", or wants to persist something.
Audit a codebase for dead code, orphaned modules, unused dependencies, and unwired seams. Quick tier runs knip-class checks ephemerally (no repo mutation); deep and exhaustive tiers add intent↔code reconciliation. Portable to any repo, including ones with no tooling installed.
Define a project's WHY/WHAT foundation autonomously — the DIRECTION entry point. Drives the definitional steps to a coherent bar via the engine (autonomous-propose): grounds every field in a cited signal, asks only the few high-leverage decisions. Triggers: 'start a project', 'define direction', 'kick off', 'set up…
Process user feedback into investigated bug specs with root cause analysis. Triggers: 'triage feedback', 'process bug reports', 'user testing feedback'.
Core automation engine. Drives a set of steps to a confirmed ground-truth bar via a main-session delegate → reconcile → re-dispatch loop. Phase skills (build, review, direction) call it; can also run directly. Two modes: autonomous (run to done unattended) and steered (coordinate, surface to user). Triggers: invoked…
Aggregates project context and audits production readiness against a comprehensive SaaS checklist. Routes to prerequisite skills when context is missing.
Define what to build — features, UX patterns, and differentiators that solve discovered user needs. Triggers: 'product definition', 'what to build', 'features', 'solution design'.
Visual QA using Playwright as a browser automation library. --setup installs Playwright and creates ui-goals.json; default mode writes inline scripts to navigate, screenshot, and report against goals.
Infrastructure skill — a single-resource advocate in the thinking-partner adversarial debate. Dispatched (not user-invoked) by shipkit-thinking-partner's adversarial mode, once per advocate per round.