durable-files-weekly-review-public

What This Skill Does

The durable-files-weekly-review-public skill is an essential utility for OpenClaw users who maintain complex agent environments. By auditing your primary instruction files (such as AGENTS.md, SOUL.md, and MEMORY.md), it monitors token usage and identifies obsolete information or redundant data. Instead of making autonomous changes, the skill generates a structured markdown report that summarizes findings and provides an approval-gated queue for cleanup actions. It ensures your agent workspace remains performant and lean without the risk of losing critical configuration history.

Installation

To install this skill, use the following command in your terminal within the OpenClaw environment: clawhub install openclaw/skills/skills/andyylin/durable-files-weekly-review-public

Use Cases

• Token Management: Prevent context window overflow in long-running agent sessions by identifying oversized configuration files.
• Maintenance: Regularly prune outdated user preferences or legacy tool definitions that no longer serve a purpose.
• Workspace Hygiene: Automate the auditing process for large teams where multiple contributors might inadvertently bloat core instruction files.
• Audit Trails: Generate dated reports that document what content was reviewed and which sections were proposed for removal before any changes are applied.

Example Prompts

1. "Run the weekly durable file audit for this workspace and show me the top three files taking up the most tokens."
2. "Can you perform a check on my AGENTS.md and MEMORY.md files and create a cleanup proposal for anything flagged as stale?"
3. "Summarize the latest durable file review report and list the specific deletions you are suggesting for my approval."

Tips & Limitations

• Explicit Approval: This skill is designed with a 'safety-first' philosophy; it will never perform silent deletions. You must review the generated report and explicitly confirm any proposed actions.
• Scope: By default, it targets a standard set of workspace files. You can point it to custom directories using the --root argument, but ensure the files are in standard markdown format for the parser to function correctly.
• Frequency: We recommend running this on a weekly basis to keep token usage within optimal limits for your specific AI model's context window.