Agent
End-user simulation specialist for overnight exploration. Tests actual usage scenarios, checks if things work as expected, identifies UX friction, broken flows, and confusing behavior. Returns structured JSON report.
Command
Last updated: 2026-07-24T14:19:53Z Total entries: 18 Convention: kebab.
Command
Single-use break-glass for a declared grant-aware safety hook; it never overrides a settings DENY. Requires an explicit narrow selector. Forms — /allow (literal, upgraded to regex only when it contains true regex metacharacters), /allow --tool (always literal, regex off), or /allow re: (explicit regex, must be…
Command
Snapshot current progress to a refs/checkpoints/ safety ref — the branch HEAD is never moved and your working branch is never pushed; the checkpoint ref itself is background-pushed (rate-limited to 1/30s per repo, without -f).
Command
Aggressive project cleanup - normalize docs structure, archive everything, delete one-time scripts/tests. Pass --codex to enable adversarial codex consultation on cleanliness-inspector and style-inspector; default is self-review only.
Command
Close the current dev cycle (agent infers task-id from conversation). QA evaluates Workflow Integrity bullets and returns CLOSE YES/NO. Pass --codex to enable multi-round QA-codex debate; default is QA-only single-round assessment. Append --force to skip the debate entirely.
Command
Delegate a task to OpenAI Codex CLI (gpt-5.6-sol, xhigh reasoning) for a second opinion or parallel coding.
Command
Commit session changes via changelog-analyst subagent.
Command
Enhanced development workflow with BA subagent delegation, command development best practices, Three-Party Architecture, and comprehensive automation patterns.
Command
Autonomous overnight development loop - continuously explores codebase, finds issues, fixes them, and repeats until end-time.
Command
Orchestrated development workflow with BA subagent delegation, parallel agent execution, and iterative QA verification. Pass --codex to enable adversarial codex consultation on each subagent's draft; default is self-review only.
Command
Allow main agent to bypass orchestrator-gate restrictions for this turn (subagent-only operations become directly allowed). Auto-clears at stop.
Command
Merge the current overnight worktree branch into the default branch (agent infers branch from active overnight state). Bare /merge typical; explicit /merge overrides. Auto-cleans worktree + branch + overnight-state file when merge succeeds and the diff is clean.
Command
Execute safe git pull with automatic stash management and conflict detection.
Command
/push is the validated wrapper for normal branch publication. Note: pretool-git-privilege-guard.py is REGISTERED in settings.json (PreToolUse, Bash matcher) and enforces commit authorization — agents must hold a valid commit grant or use the auto-bulk: bridge prefix to commit. The wrapper script /.claude/hooks/push.sh…
Command
Run the /dev workflow with the full command specification already loaded in this conversation.
Command
Resume every quota-interrupted subagent in the current Claude Code parent session from its original transcript and agent ID.
Command
Continuation spec update or temp session note (was /update then /spec-continue — renamed to avoid collision with MAP's /update portfolio mutation command).
Command
Create spec files for any dev workflow (/dev, /dev-overnight, or standalone reference). Pass --codex to enable adversarial codex consultation on each spec-subagent / QA dispatch; default is self-review only.
Command
Cancel active overnight time-lock + workflow-enforce so the session can terminate normally. User-invoked only — agents cannot self-stop.
Command
Test validation workflow with edge case detection, systematic validation, and quality enforcement.
Skill Claude CodeCodex
Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.
Skill Claude CodeCodex
Apply the 58-rule anti-pattern catalog (10 Color + 5 Motion + 5 Typography + 5 Spacing + 2 Glass + 5 Heuristic + 4 UX-Writing + 5 Form + 4 Interactive + 5 Nielsen + 8 AI-slop) against a Playwright page. Outputs aestheticfindings[] with category=harddefect|tasteheuristic, with the SCHEMA-ENFORCED severity hard-cap on…
Skill Claude CodeCodex
Run APCA Lc text-contrast measurement on a Playwright page in BOTH light and dark color schemes. Returns deterministic apca. findings against rule-map.json. Use during ui-specialist Phase 6 (Accessibility).