| name | tooling-registry |
| description | Map Hara Vital's Codex commands, including start-session/end-session, audits, bug hunts, feature starts, PR prep, and test scaffolding, so they trigger automatically from chat requests. |
Hara Codex Tooling Registry
Use this skill whenever a request matches any of the command phrases below or when a plan is required. It links natural-language triggers to the scripts/command docs we added under .codex/ so Codex can respond like Claude.
Session workflows are part of the primary trigger surface. If the user asks to start or end a session, run those workflows before falling back to a generic repo bootstrap.
Command Triggers → Actions
| Trigger phrases | Action |
|---|
/start-session, "start session", "run start-session" | Follow .codex/commands/start-session/COMMAND.md: read the active plan (default .claude/plans/main.md), run node scripts/codex/git-session-summary.mjs "2 days ago", then brief context + next steps + blockers. |
/end-session, "end session", "run end-session" | Follow .codex/commands/end-session/COMMAND.md: run node scripts/codex/git-session-summary.mjs "today 00:00", append the new ### Session — YYYY-MM-DD block to the active plan, and report carryover items. |
/audit, "run audit on ", "complexity report" | Run node scripts/codex/audit.mjs <path>; include Markdown report from .codex/commands/audit/COMMAND.md. |
/bug-hunt, "find bugs in " | Follow .codex/commands/bug-hunt/COMMAND.md checklist, covering React/TS/Supabase patterns. |
/feature-start, "kick off feature", "plan feature" | Run through .codex/commands/feature-start/COMMAND.md (analysis, reuse check, checklist). |
/pr-prep, "prepare PR" | Execute node scripts/codex/pr-prep.mjs (build/lint/test + scans) and finish the manual checklist. |
/test-create, "add tests for " | Use .codex/commands/test-create/COMMAND.md templates for components/hooks/utils/API routes. |
When a trigger is detected, explicitly mention which command is running and summarize the results/output.
Planning Workflow
How to Use
- Detect whether the user’s request matches a trigger phrase or requires planning.
- Load the corresponding command instructions and run the script if one exists.
- Share outputs (audit table, scan findings, plan summary, etc.) back to the user.
- When multiple commands are requested, execute them sequentially, reporting after each.