excatt Command Reference
The authoritative skill list (names + descriptions) is auto-injected by the harness every session — enumerate that when the user asks what is available. This reference only groups the framework's own skills by role. Information display only — no execution.
Framework Core (workflow mechanisms)
| Skill | Purpose |
|---|---|
/confidence-check |
Pre-implementation confidence assessment (Medium+ difficulty) |
/verify |
6-phase verification (build → type → lint → test → security → diff) |
/checkpoint |
Git-based restore point before risky operations |
/tdd |
RED-GREEN-REFACTOR cycle enforcement |
/build-fix |
Resolve build errors with minimal changes |
/audit |
Project-specific rule validation |
/gap-analysis |
Design-vs-implementation Match Rate (PDCA Check) |
/feature-planner |
Phase-based feature planning |
/learn |
Extract and save session patterns |
/note |
Persist memos across sessions |
/fix-pr |
Collect PR review comments and fix |
/config-doctor |
Framework configuration diagnosis |
/eval-harness |
Eval-driven development for AI features |
Review & Critique
/devils-advocate (decision challenge) · /business-panel (expert panel) ·
/brainstorm (requirements discovery) · /grill-with-docs (domain model
stress-test) · /security-audit · /react-best-practices ·
/python-best-practices · /composition-patterns · /web-design-guidelines
Design & Frontend
/ui-ux-pro-max (design intelligence) · /frontend-design ·
/impeccable + 17 sub-commands (/shape, /layout, /typeset, /colorize,
/animate, /delight, /polish, /critique, /design-audit, /harden,
/optimize, /clarify, /distill, /quieter, /bolder, /adapt,
/overdrive) · /theme-factory · /brand-guidelines · /canvas-design ·
/algorithmic-art
Documents & Tooling
/pdf · /pptx · /xlsx · /docx · /internal-comms · /artifacts-builder ·
/slack-gif-creator · /mcp-builder · /skill-creator · /webapp-testing ·
/agent-browser · /pytest-runner · /uv-package
Removed in v3.0
Generic topic guides (architecture, caching, docker, GraphQL, naming, ...) and thin command wrappers (analyze, implement, improve, cleanup, ...) were removed — the model handles these natively. Ask directly instead of looking for a skill.