grtninja/skill-hub

Skill Hub

Use this skill as the entry router for skill chaining.

Workflow

1. Parse request scope (repositories, domains, risk level, and deliverables).
2. For meta-harness or large-system work, capture the current local authority contract first:
   • treat G:\GitHub as the canonical repo root on the maintainer workstation
   • treat http://127.0.0.1:9000/v1 and http://127.0.0.1:2337/v1 as authoritative model lanes
   • treat http://127.0.0.1:1234/v1 only as a non-authoritative operator surface
3. Select skills from the routing matrix below and keep the chain to the minimum set that fully covers the request.
4. Order the chain as: routing -> guardrails -> execution -> release/policy gates.
5. Emit or capture the chain output contract before execution begins.
6. Add loopback criteria for unresolved lanes before execution starts.

Routing Matrix

| Scenario | Example Request | Chain Pattern | | -------- | --------------- | ------------- | | Single-repo, low risk | "Fix a bug in the admin terminal" | skill-hub -> repo-a-host-admin-ops -> skill-common-sense-engineering | | Multi-repo, policy risk | "Sync packaging across repo-A and repo-D" | skill-hub -> usage-watcher + skill-cost-credit-governor -> execution skills -> skill-enforcer | | New or changed skills | "Create a new overlay skill for Blender" | skill-hub -> guardrails -> skill-creator-openclaw -> skill-auditor -> skill-arbiter-lockdown-admission | | Interrupted work | "Continue the release prep from yesterday" | skill-hub -> request-loopback-resume -> remaining lanes | | Independent lanes | "Update docs AND run host validation" | split into parallel chains; merge only after per-lane evidence is present |

Chain Rules

• Prefer one primary execution skill per independent lane.
• Avoid optional skills unless they add deterministic value for the request.
• For multi-repo work, include a policy-enforcement lane before finalizing code changes.
• For new or updated skills, include auditing and lockdown-admission lanes.
• Prefer healthy local OpenClaw-compatible subagents first for quick bounded sidecar work.
• When cloud sidecars are unavoidable, keep them on a lower-reasoning, low-cost sidecar lane.
• For whole-PC or meta-harness tasks, capture PC Control local-agent or status-surface evidence before assuming missing context.
• If a required skill is missing, fail closed with a bounded fallback and record the gap.

Chain Output Contract

Emit or capture a compact chain payload before execution:

• task
• lanes[] (ordered lane identifiers)
• skills_by_lane
• evidence_paths[] (usage/cost/cold-warm + audit/arbiter when applicable)
• loopback_criteria
• stop_conditions (done, blocked, reroute)

Illustrative example:

{
  "task": "update host packaging and add new overlay skill",
  "lanes": ["host-packaging", "skill-creation"],
  "skills_by_lane": {
    "host-packaging": ["repo-a-host-admin-ops", "skill-common-sense-engineering"],
    "skill-creation": ["skill-creator-openclaw", "skill-auditor", "skill-arbiter-lockdown-admission"]
  },
  "evidence_paths": [
    "/tmp/usage-analysis.json",
    "/tmp/usage-plan.json",
    "/tmp/skill-cost-analysis.json",
    "/tmp/skill-cost-policy.json",
    "/tmp/cold-warm-analysis.json",
    "/tmp/cold-warm-plan.json"
  ],
  "loopback_criteria": "re-route if host validation fails or audit blocks admission",
  "stop_conditions": ["all lanes emit evidence", "blocked lane triggers loopback", "operator cancels"]
}

If this contract is incomplete, chain selection is incomplete and must fail closed.

Referenced Skills

Key skills this hub routes to (see each skill's SKILL.md for full details):

• Guardrails: usage-watcher, skill-cost-credit-governor, skill-cold-start-warm-path-optimizer
• Execution: repo-a-host-admin-ops, repo-b-* skills, skill-creator-openclaw, domain-specific skills
• Policy gates: skill-enforcer, skill-arbiter-lockdown-admission, skill-auditor
• Resume: request-loopback-resume
• Hygiene: skill-common-sense-engineering

Scope Boundary

Use this skill only for request-to-chain routing and loopback decisioning.

Do not use this skill to implement repository changes directly.

Loopback

If any lane remains unresolved, blocked, or ambiguous:

1. Capture lane blockers and current evidence.
2. Re-run routing through $skill-hub with updated constraints.
3. Resume only when every lane has a deterministic next action.