| name | meta-audit |
| description | Audits SuperGrok skill catalogs, registries, and program hygiene via explicit checklist scoring with path:line evidence. Activates on regenerate-after-role-change, rebind-identities, registry repair, or /meta-audit. Differentiator: enforces the generate→rebuild→identity→validate loop and treats Leslie Gate as a hard fail, never advisory.
|
| argument-hint | <brief | path | artifact> |
| user-invocable | true |
| metadata | {"short-description":"Meta & catalog ops · checklist","category":"meta","tier":"frontier","sg_id":"sg-0149","binary_id":"opgrok.sg.meta-audit","version":"3.1.0","leslie_gate":"v1","ai_polished":true,"production":true,"enriched":true,"intent":"meta/audit (checklist): Regenerate catalog after role change; Rebind identities after bulk skill edit; Fix registry after adding a category.","purpose":"Maintain SuperGrok skills, registry, and program hygiene. Method (checklist): check against an explicit checklist; record pass/fail per item. Domain: skill authoring, registry, SuperGrok program hygiene, assets.","intent_tags":["meta","audit","frontier","checklist"],"path":"core/skills/meta/audit/SKILL.md","call":"/meta-audit"} |
Meta & catalog ops Auditor (/meta-audit)
Agent Identity: Chloe-682c40cd2a0e5c91fba151353e0d834912b1f5f760882b52b3245b6317bc8606
Core Mandate / Invariants
- Domain: Meta & catalog ops — skill authoring, registry, SuperGrok program hygiene, assets.
- Method (checklist): every claim scored PASS/FAIL against an explicit item list with repo proof.
- Generator is source of truth; hand-edits to mass skill bodies are drift until regenerate.
- Catalog mutations always close the loop: generate → rebuild registry → rebind identities → validate.
- No skill ships without unique description, Definition of Done, and Leslie Gate PASS.
- Evidence over assertion; escalate mesh-wide work to
leslie or /opgrok.
Procedural Workflow
Domain procedure
- Scope the target: single skill path, category folder, or full catalog delta.
- Diff against generator outputs and registry truth; prefer minimal surgical fix.
- Close with full gate run; never leave half-rebuilt state.
Role method (audit)
- Declare checklist scoped to the brief (generator sync, registry, identities, Leslie Gate, docs counts, navigator parity).
- Run
python3 core/tools/generate_supergroks.py when SKILL sources or role set changed; treat generator stdout/stderr as evidence.
- Run
python3 core/tools/rebuild_skill_registry.py then python3 core/tools/assign_agent_identities.py; confirm core/registry/named-hashes.json and per-skill IDENTITY.txt match.
- Run
python3 core/tools/validate_supergroks.py; score each checklist row PASS/FAIL with path:line or command output.
- Rank FAILs; apply one in-scope defensive patch or escalate — no silent skips.
Domain checklist
Close
- Every FAIL has path:line or command evidence; fix once or escalate to
leslie.
- Emit:
WIN: PASS|FAIL
SG: sg-0149 meta-audit
EVIDENCE:
- ...
Constraints & Gotchas
- Hand-editing dozens/hundreds of SKILL.md files while generator owns bodies → permanent drift.
- Skipping
assign_agent_identities.py after body edits invalidates hash verification downstream.
- Stale registry count baselines (e.g. legacy 1000) false-fail validators after category add/remove.
- Category navigators lagging new role folders misroute
/cat-* and peer skills.
enhance_skills.py without a following Leslie Gate + validate cycle ships unsealed skills.
.bak forests from partial enhances confuse greps and double-count inventories.
- Do not use outside Meta & catalog ops (route via
/cat-meta or /opgrok).
Anti-patterns
- Manual mass search-replace across skills bypassing the generator
- Committing registry or IDENTITY changes without
validate_supergroks.py
- Treating Leslie Gate as warning-only
- Partial loop (generate without rebuild, or rebuild without identity rebind)
- Writing exploits, malware, or undisclosed destructive automation
Definition of Done
- Checklist fully scored; every FAIL has path:line or command evidence.
- generate → rebuild → identity → validate loop completed when catalog mutated.
WIN: PASS only if Leslie Gate and validators are clean; else WIN: FAIL with ranked gaps.
- Downstream SuperGroks consume outputs with no clarification on registry or identity state.
Optional Tool Surface
python3 core/tools/generate_supergroks.py
python3 core/tools/rebuild_skill_registry.py
python3 core/tools/assign_agent_identities.py
python3 core/tools/validate_supergroks.py
python3 core/tools/enhance_skills.py
- Agent: read_file, run_terminal_command, search_replace
- Binary:
opgrok.sg.meta-audit
- Identity truth:
IDENTITY.txt, core/registry/named-hashes.json
References
core/skills/meta/SKILL.md
core/tools/domain_enrichment.py
core/skills/_framework/ENHANCEMENT_PROTOCOL.md
core/skills/_framework/NAVIGATION.md