Skill router
A lightweight Codex meta-skill for skill routing, Goal/Plan mode selection, and read-only-first safety checks.
npx -y skills add wpq250dollars/codex-skill-router --skill skill-routerAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
skill selection, goal, plan, high-risk task detection, browser automation, login workflow, config changes, GitHub operations, database work, project scaffolding, multi-skill workflow. Use to decide which Codex skills to use, whether Goal or Plan mode is appropriate, whether read-only inspection is needed, and how to chain multiple skills safely.
SKILL.md
4.0 KB, as published. Nobody here has run it
Skill Router
Use this skill as a default routing layer before substantial Codex work. Keep routing silent for simple low-risk tasks. Show a routing summary only when the task is complex, ambiguous, multi-skill, or high-risk.
Core Decisions
Decide these before acting:
- Relevant skills to use.
- Goal mode or Plan mode.
- Whether multiple skills should be chained.
- Whether read-only inspection must happen first.
- Whether the task contains high-risk operations.
Silent vs Visible Routing
Use silent routing when the task is simple, low-risk, and clearly scoped, such as small edits, explanations, one-file bugs, or direct questions.
Show a short routing summary when any of these apply:
- New project, architecture, scaffolding, or multi-file implementation.
- Browser automation, login, account/session workflow, or screenshots.
- Database, Supabase, migration, auth, RLS, production data, or schema changes.
- GitHub PR, issue, CI, push, release, publish, or deployment.
- Config, environment variables, secrets, credentials, or system settings.
- Deleting files, moving many files, cleanup, reset, overwrite, or destructive operations.
- Multiple skills are likely needed.
- The user explicitly asks for planning, safety boundaries, or skill selection.
Mode Selection
Recommend Goal mode for low-risk work:
- Small clear fixes.
- Single-file or narrow changes.
- Clear bug with direct reproduction.
- No account, browser, database, config, GitHub write, deletion, or release risk.
Recommend Plan mode for higher-risk or broader work:
- New project or multi-file design.
- Architecture decisions.
- Browser automation, login, or account actions.
- Database work, migrations, Supabase, RLS, auth, or production data.
- Config changes, environment variables, secrets, credentials, or tool setup.
- GitHub operations, CI debugging, PR review, push, release, or deployment.
- File deletion, cleanup, reset, bulk moves, or overwrite operations.
If a task includes deletion, config/env changes, login, browser automation, database work, GitHub push/release, or deployment, recommend Plan mode and start with read-only inspection.
Complex/high-risk tasks must stop after plan/read-only inspection and wait for explicit user approval before write actions.
Skill Matching
Prefer discovered skills from data/skills_cache.json when available. Use data/custom_rules.json to improve unknown or local skills. If the cache is missing or stale, ask to run the refresh script from the installed skill directory:
python skill-router/scripts/refresh_skills.py
Use common pairings:
- PDF to slides:
pdfthenpresentations. - Spreadsheet or CSV analysis:
spreadsheetsand, for analytical decisions or dashboards,data-analytics. - GitHub PR or CI:
github, thengh-fix-ciorgh-address-commentswhen relevant. - OpenAI API app:
openai-platform-api-key, thenopenai-docsor OpenAI developer skills. - Supabase or Postgres:
supabase, thensupabase-postgres-best-practicesfor SQL/schema/performance. - Browser testing or local web UI:
browserorplaywright-interactive. - Desktop app automation:
computer-useorelectronwhen applicable.
If a discovered skill has no registry/custom explanation, mark it as unknown, keep it visible as a candidate, and suggest adding metadata to data/custom_rules.json.
Visible Output Template
For complex or high-risk work, keep the summary short:
Routing:
- Mode: Plan
- Skills: skill-a -> skill-b
- Read-only first: yes
- Risk flags: config change, browser automation
- Boundary: inspect first; do not modify/delete/push until plan is confirmed
Then continue according to the active collaboration mode and user instructions.