Skills Third-Party Intake
Use this skill to turn external skill catalogs into a safe, ranked intake list.
Workflow
- Scan one or more third-party skill roots.
- Run deterministic checks inspired by hardened loaders:
- frontmatter/schema hygiene
- script/link risk markers
- high-risk snippet detection
- dependency/sensitive marker inventory
- Score each skill for compatibility, quality, and security.
- Produce an
admit/manual_review/rejectrecommendation list. - Feed only
admitcandidates intoskill-installer-plus+skill-arbiterlockdown admission.
Intake Scan Command
python3 "$CODEX_HOME/skills/skills-third-party-intake/scripts/third_party_skill_intake.py" \
--source-root openclaw=G:/_3rdParty-Clones/openclaw/skills \
--source-root openclaw-ext=G:/_3rdParty-Clones/openclaw/extensions \
--json-out /tmp/third-party-intake.json
Review-Only Mode
python3 "$CODEX_HOME/skills/skills-third-party-intake/scripts/third_party_skill_intake.py" \
--source-root /path/to/external/skills \
--min-security 0.75 \
--format table
Admission Handoff
python3 "$CODEX_HOME/skills/skill-arbiter/scripts/arbitrate_skills.py" \
<skill> [<skill> ...] \
--source-dir skill-candidates \
--window 10 --baseline-window 3 --threshold 3 --max-rg 3 \
--personal-lockdown \
--json-out /tmp/skills-third-party-intake-arbiter.json
Guardrails
- Fail closed on high-risk snippets and unsafe markdown links.
- Treat script-bearing third-party skills as manual review unless explicitly justified.
- Keep imports public-shape only; strip private paths and identifiers before commit.
Scope Boundary
Use this skill only for third-party discovery and intake triage.
Do not install directly from third-party repos in this lane; route admitted candidates through skill-arbiter lockdown mode.
References
references/intake-rubric.md