From claude-code-hermit
Lints raw/ and compiled/ knowledge directories for stale, unreferenced, missing-type, and oversized artifacts. Runs Node.js script for read-only reports with actionable advice. Activates on 'check knowledge', 'lint knowledge', 'knowledge health'.
npx claudepluginhub gtapps/claude-code-hermit --plugin claude-code-homeassistant-hermitThis skill uses the workspace's default tool permissions.
Read-only lint of the hermit's knowledge directories. Reports findings without modifying any files.
Audits and interactively cleans up Claude Code knowledge base: detects stale references, duplicates, orphaned files, frontmatter issues, and merge opportunities.
Audits knowledge graph for orphan nodes, missing cross-references, stale claims, unresolved contradictions, low-confidence clusters, and code-knowledge gaps. Outputs health report with auto-fixes.
Verifies and updates existing codebase documentation nodes (docs/*.md) against current code after changes or refactors. Fast mode for triaged items; full audits entire knowledge base.
Share bugs, ideas, or general feedback.
Read-only lint of the hermit's knowledge directories. Reports findings without modifying any files.
/claude-code-hermit:knowledge — lint (default)
/claude-code-hermit:knowledge lint — same
Run the shared lint script:
node scripts/knowledge-lint.js .claude-code-hermit
The script path is relative to the plugin install directory. Use Bash to execute it.
Present the script output to the operator as-is — it already formats findings grouped by type with file paths, ages, and actionable advice.
If the script exits cleanly with "Knowledge base is clean", relay that message. If it reports findings, relay the full output.
Never touches: OPERATOR.md, knowledge-schema.md, config.json, proposals/, sessions/, state/
This skill is strictly read-only. It does not move, delete, or modify any files.