Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
- Clarified security and privacy sections to emphasize that the skill's Node.js code is fully local; outbound data depends on agent/model stack used. - Added details about optional environment variables for easier customization and clearer configuration. - Explained implications of included agent config templates (OpenAI, OpenRouter): prompts may leave the machine if run via certain agents. - Enhanced documentation of master directive behavior, especially mutation forcing and safe usage flags. - Removed redundant documentation files (CHANGELOG.md, CLAWHUB_UPLOAD.md).
What This Skill Does
Audits OpenClaw session logs, workspace memory, and installed skills to identify breakages and value leaks. Runs structured mutation cycles that select a repair or improvement mode, apply changes, and persist evolution state for compounding gains. Supports single-shot, review-gated, dry-run, and continuous loop execution modes.
Combines log inspection, memory reading, and skill mutation into a single parameterized cycle with compounding persistent state, rather than requiring manual correlation of logs, memory, and skill files separately.
When to use it
- Diagnosing repeated agent failures from session logs
- Running continuous background evolution loops overnight
- Reviewing proposed mutations before they modify workspace files
- Generating improvement prompts without applying any changes
- Tracking capability changes across cycles with persistent state
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: run an evolution cycle on my main agent
- 1Reads recent session logs from ~/.openclaw/agents/main/sessions/*.jsonl up to TARGET_SESSION_BYTES limit
- 2Loads workspace memory from MEMORY.md, memory/YYYY-MM-DD.md, and USER.md
- 3Selects mutation mode (repair, optimize, expand, instrument, or personalize) based on detected breakages and value leaks
- 4Writes mutation instructions and a cycle report with identified issues
- 5Persists evolution state to memory/evolution_state.json and long-term learning to memory/funky_fund_flamingo_persistent_memory.json
Cycle report listing detected issues, selected mutation mode, and applied or recommended changes with ROI rationale