Skip to content

SKILL.md

Latest commit

 

History

History
38 lines (31 loc) · 1.73 KB

SKILL.md

File metadata and controls

38 lines (31 loc) · 1.73 KB
name memory-dream-governance
description Memory dream governance for OpenClaw memory stack. Trigger this skill whenever user asks to “dream”, “auto dream”, “memory governance”, “记忆治理”, “记忆整理”, “整理 MEMORY.md / memory/*.md”, or wants dry-run/audit/apply/rollback workflow on memory files with safe lock and explicit apply gate.

Memory Dream Governance

Use this skill to run a safe, auditable memory-governance pipeline for workspace memory assets.

Triggers / 触发词

  • English: dream, auto dream, memory governance, consolidate memory, memory cleanup, audit gate
  • 中文: 记忆治理, 记忆整理, 梦境整理, 整理 MEMORY, 审计放行, 回滚记忆改动

Scope & Safety

  • Write scope: MEMORY.md, memory/**/*.md, memory/**/*index*.md
  • Never write: SOUL.md, AGENTS.md, TOOLS.md, openclaw.json, source code files
  • Default mode: dry-run only
  • Apply requires explicit --apply
  • Lock file required to avoid overlapping runs

Minimal flow

  1. python3 scripts/dry_run.py --workspace ../../
  2. python3 scripts/audit.py --plan <plan.json>
  3. Optional apply (manual gate only): python3 scripts/apply.py --plan <plan.json> --audit <audit.json> --apply
  4. Rollback if needed: python3 scripts/rollback.py --snapshot <snapshot.json> --apply

Output artifacts

  • artifacts/plans/plan-<run_id>.json
  • artifacts/audit/audit-<run_id>.json
  • artifacts/snapshots/snapshot-<run_id>.json
  • artifacts/reports/report-<run_id>.md

Operational notes

  • Auto apply is disabled by default (config/default-policy.yaml)
  • Wrapper script: ../../scripts/cron-memory-dream-governance.sh
  • If audit result is REVIEW or BLOCK, do not apply automatically.