quest

Evaluates the accuracy and scope coverage of the other audit skills (perf-audit, test-audit, doc-audit, wiki-audit, dependency-audit) by re-verifying past findings against source and checking for scope drift, then safely improves their SKILL.md files. Use when asked to evaluate audit skill performance, or auto-triggered by a domain audit skill after its 5th run since the last evaluation.

Multi-agent dependency health audit — finds outdated versions, unused deps, security issues, and over-specified features. Use when dependencies are stale, after adding deps, or before releases.

Multi-agent wiki audit — finds stale numbers, missing systems, and broken links in the player-facing wiki. Use when wiki is stale, after game changes, or before releases.

Multi-agent developer documentation audit — finds stale constants, missing files, and outdated types across CLAUDE.md files, README.md, and docs/. Use when docs are stale, after adding features, or before releases.

Multi-agent test health audit — finds flaky tests and performance bottlenecks by area, auto-fixes safe patterns, and verifies with 10x runs. Use when tests are flaky, slow, or before releases.

Drive the real game in tmux against fixture save states, inspect PNG screenshots or short video recordings of the rendered TUI to verify your own UI changes, and iterate until they render correctly. Use PROACTIVELY after any change to src/ui/, when asked to "screenshot the game", "record the game", "show me the change/transition/animation", "drive the game", or before opening a PR that touches UI.

Write a new docs/dossiers/*.md design dossier or refresh an existing one — a living, cross-system, player-eye synthesis of an Act or major system, grounded in source citations and simulator evidence. Use when asked to create a dossier, refresh a dossier, or check whether a dossier reflects the current design.

Multi-agent performance audit — finds and fixes hot-path bottlenecks, adds criterion benchmarks and simulator profiling. Use when game feels slow, after adding features, or to establish performance baselines.

Multi-agent game balance simulator — runs the headless simulator across strategies and seeds to test Z50/Ascension X progression, then produces a prioritized balance report. Use when you want to check game balance, after landing combat/zone/loom changes, before releases, or when asked to "run the simulator", "check balance", or "can we beat Z50".

Complete checklist for adding a new challenge minigame to Quest. Use this skill whenever someone wants to add a new challenge, minigame, puzzle, or game — even if they don't say "challenge" explicitly. Covers all 15 integration points across the codebase so nothing gets missed. Use PROACTIVELY when implementing any new challenge from a spec or plan.

Run all audit skills — perf-audit, test-audit, doc-audit, wiki-audit, dependency-audit — in parallel on isolated worktrees. Use when you want a full project health check or before a release.

Push branch, create PR with automerge, and watch until merged. Only use when explicitly asked to "ship" or invoked via /ship. Do NOT trigger on generic phrases like "push", "merge", or "land".

Use when asked to clean up, reset workspace, or make repo look like a fresh clone. Use after finishing feature work, merging PRs, or when workspace has stale branches and uncommitted files.