Agent Skills
› NeverSight/learn-skills.dev
› gov_audit
gov_audit
GitHub用于在引导、迁移或应用后执行治理完整性审计。检查清单一致性、锚点、路径兼容性及系统证据，验证平台控制面变更备份和脑文档修改流程，确保官方流兼容性并输出标准化结果。
Trigger Scenarios
引导完成后
数据迁移后
配置应用后
Install
npx skills add NeverSight/learn-skills.dev --skill gov_audit -g -y
SKILL.md
Frontmatter
{
"name": "gov_audit",
"metadata": {
"openclaw": {
"emoji": "✅"
}
},
"description": "Run post-bootstrap or post-migration governance audit.",
"user-invocable": true
}
/gov_audit
Purpose
Perform governance integrity checks after bootstrap, migration, or apply.
Required checks
- Run the checklist in
_control/REGRESSION_CHECK.mdwith fixed denominator 12/12. - Verify governance anchor consistency required by your active migration baseline.
- Produce a clear PASS/FAIL result and remediation if any item fails.
- Verify path compatibility:
- governance content must use runtime
<workspace-root>semantics - no hardcoded
~/.openclaw/workspaceassumptions in changed governance content
- governance content must use runtime
- Verify system-truth evidence:
- OpenClaw system claims must cite
https://docs.openclaw.aisources - latest/version-sensitive OpenClaw claims must also cite
https://github.com/openclaw/openclaw/releasessources - date/time claims must include runtime current time evidence (session status)
- OpenClaw system claims must cite
- If a run includes platform control-plane changes, verify:
- backup path exists under
archive/_platform_backup_<ts>/... - before/after key excerpts are present
- change was executed via
gov_openclaw_jsonpath (or equivalent documented fallback)
- backup path exists under
- If a run touches Brain Docs (
USER.md,IDENTITY.md,TOOLS.md,SOUL.md,MEMORY.md,HEARTBEAT.md,memory/*.md), verify run report includes:FILES_READexact pathsTARGET_FILES_TO_CHANGEexact paths (ornonefor read-only) Missing either field => FAIL (evidence incomplete).
- If a run includes coding/workspace file writes (for example under
projects/), verify it was treated as Mode C with:- explicit PLAN gate evidence
- READ evidence
- QC 12/12 outcome Missing evidence => FAIL (workflow bypass).
- If a run includes
gov_brain_audit APPROVE: ...orgov_brain_audit ROLLBACK, verify:- backup path exists under
archive/_brain_docs_autofix_<ts>/... - run report includes approved finding IDs (or rollback scope)
- changed files are limited to approved Brain Docs/governance targets Missing evidence => FAIL (unsafe Brain Docs mutation).
- backup path exists under
- Official-flow compatibility SOP check (anti-self-lock):
- verify governance does not falsely block OpenClaw system operation flows (
openclaw ..., including plugin-added/future commands) and governance lifecycle flows (gov_help,gov_setup quick/check/install/upgrade,gov_migrate,gov_audit,gov_openclaw_json,gov_brain_audit,gov_uninstall quick/check/uninstall). - if a governance block occurred, verify run report labels it as governance policy gate (not system error) and provides copy-paste unblock commands.
Persistence
- Write audit result into
_runs/when the active governance flow requires persistence. - Ensure
_control/WORKSPACE_INDEX.mdis updated when a new run report is added.
Output requirements
- Use this output order for UX consistency:
STATUSWHYNEXT STEP (Operator)COMMAND TO COPY
- Always include a final
NEXT STEP (Operator)section. - If audit PASS:
- primary: continue normal operation, or run
/gov_apply <NN>only when an approved BOOT menu item exists. - fallback:
/skill gov_apply <NN>
- primary: continue normal operation, or run
- If audit FAIL:
- primary: run
/gov_migrateafter remediation. - fallback:
/skill gov_migrate
- primary: run
Fallback
- If slash command is unavailable or name-collided, use:
/skill gov_audit
Version History
- e0220ca Current 2026-07-05 23:04
