self-healing

Pass

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell scripts (detect-failure.sh, new-heal.sh, find-similar-heals.sh) to automate the workflow of identifying non-zero exit codes from other tools and logging them to .learnings/HEALS.md. These scripts perform standard filesystem operations (mkdir, touch, cat, grep) and use Python for basic JSON parsing and regex matching on local files. This is standard behavior for developer-focused agent skills.
  • [REMOTE_CODE_EXECUTION]: While the documentation mentions external services and APIs, it specifically instructs the agent to 'patch the call shape' or 'find a workaround' locally rather than downloading and executing arbitrary remote payloads. The provided examples show local fixes like switching package managers (npm to pnpm) or updating environment settings (nvm use).
  • [SAFE]: References to external repositories like browser-use/browser-harness are provided purely for conceptual documentation and inspiration, with no automated downloads or execution from these sources.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 12, 2026, 11:37 PM
Security Audit — agent-trust-hub — self-healing