Use when agents need a bulk health-check — scans the local agents folder, runs every agent in parallel to self-verify its data is still accurate, and removes or updates any stale references
Over time, agent definition files in the agents/ folder can drift from reality — files they reference get moved, projects change structure, external resources disappear, or conventions evolve. This skill performs a bulk health-check on every agent by spawning parallel verification runs, then reconciles any stale or broken data it finds.
Use this skill whenever you want to make sure all agents in the current plugin are still accurate and functional before a release, after a major refactor, or as a periodic maintenance pass.
When this skill is invoked, scan the agents/ folder in the current project root, identify every .md agent file, and run a parallel self-verification pass on each one. For every agent, validate that:
namedescriptionname.mdAfter verification, take corrective action:
agents/ directory in the project root and collect all .md files (skip .gitkeep and non-agent files).name, description) and checks the filename matches.
c. Extracts all file paths, local paths, and URLs referenced in the agent content.
d. Verifies each file path exists by attempting to read or glob it.
e. Verifies each URL is reachable (HEAD request via WebFetch when possible).
f. Cross-references factual claims against the current codebase (e.g., if the agent describes a project structure, verify it matches).
g. Returns a structured report: agent name, status (healthy / needs-update / obsolete), list of issues found, and suggested fixes.agents/ folder of the current project root — do not scan other plugin directories or external paths.agents/ folder is empty or contains no .md files, report that clearly and exit without error.name frontmatter field.