heal-skill

To fix broken skill documentation, you can use an automated healing process that analyzes conversation context to detect errors and propose specific before-and-after diffs for your SKILL.md files. It applies the corrections after you explicitly approve the changes.

Yes, you can automatically correct code and documentation errors by reflecting on execution failures to pinpoint the running skill. The system generates targeted fixes with specific diffs and applies them through a user-guided approval workflow.

Automated skill healing works by analyzing the conversation context to identify the currently running skill, reflecting on the encountered errors, and proposing specific fixes with diffs for your documentation and code before applying changes.

Yes, you need to manually approve code corrections during the skill healing process. The system requires explicit user confirmation before applying any changes to your SKILL.md and related files, ensuring you maintain full control over the modifications.

The best way to update an incorrect API endpoint in a skill file is to use a healing skill that detects the error during execution, proposes the correct endpoint with a before-and-after diff, and applies the fix after your approval.

Diagnosing and healing a skill requires tools for file system interaction and code editing to read the documentation, generate proposed diffs, and apply the approved corrections to the SKILL.md and related files.