| name | codex-skill-universe |
| description | Operate the local Codex Skill Universe dashboard. Use when the user asks to open, start, scan, refresh, verify, troubleshoot, or maintain the Codex Skill Universe web app; inspect local Codex skills; run AI Skill Doctor; configure OpenAI API key/status; save or manage Skill Groups; view or record skill usage statistics/histograms; check /api/health, /api/skills, /api/ai/status, /api/skill-groups, or /api/skill-usage; debug Vite, Node, npm, PowerShell, port, cache, privacy, or ClawHub recommendation status for this dashboard. |
Codex Skill Universe
Operating Workflow
Use this skill to help the user run and maintain the local Codex Skill Universe web dashboard. The dashboard project is identified by a package.json whose name is codex-skill-universe.
- Locate the dashboard project. Prefer a path provided by the user. Otherwise, search the current workspace and nearby
Documents/Codex folders for the matching package.json.
- Start or refresh the dashboard with
scripts/launch.mjs. It runs npm run scan, starts the Vite dev server on 127.0.0.1, and verifies /api/health.
- Report the local URL, usually
http://127.0.0.1:5173/, plus the skill count from /api/skills when available.
- For AI Skill Doctor requests, check
/api/ai/status. If disabled, help the user configure OPENAI_API_KEY; if enabled, use the page's AI Check button or /api/ai/analyze-skill for a user-selected skill only.
- For Skill Groups, inspect
/api/skill-groups and remind the user they are local JSON files under .skill-universe/skill-groups/.
- For usage statistics, inspect
/api/skill-usage. Usage counters are local JSON under .skill-universe/skill-usage.json; record only user-confirmed usage, not private Codex logs.
- For maintenance checks, run
scripts/check.mjs. Use --build only when the user asks for a stronger release check or before publishing changes.
- If startup, scans, AI analysis, Skill Groups, usage statistics, recommendations, or privacy behavior are unclear, read
references/operator-guide.md.
Commands
From the dashboard repository:
node skills/codex-skill-universe/scripts/launch.mjs --root . --port 5173
node skills/codex-skill-universe/scripts/check.mjs --root .
node skills/codex-skill-universe/scripts/check.mjs --root . --build
When this skill is installed under ~/.codex/skills/codex-skill-universe, run the same scripts from that installed path and pass --root pointing to the dashboard repository.
Boundaries
- Do not auto-install third-party skills from recommendations. Generate reviewable install plans and ask for user approval before any install.
- Do not run AI analysis in bulk or in the background. AI Skill Doctor is user-triggered and scoped to the selected skill.
- Do not upload reference contents, assets, logs, project profiles,
.env files, local absolute paths, or secrets. AI Skill Doctor sends only selected SKILL.md frontmatter/body, scan metadata, headings, trigger terms, and resource names after redaction.
- Do not upload the full local skill catalog for AI Skill Group suggestions; related group members are added locally from the relation graph.
- The dashboard default sends only gap keywords for ClawHub recommendation search.
- Keep skill usage counters local. Do not parse or upload private Codex conversation logs to infer usage.
- Treat
.skill-universe/, .env, .env.local, logs, screenshots, and project profiles as local runtime data.