The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Bash to execute file system inspection and maintenance commands including ls, find, wc, du, rm, and git hash-object. These tools are used for diagnostic purposes such as identifying junk files for removal, calculating cache size, and verifying file integrity via hashing.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads local configuration files, including settings.json and settings.local.json, to verify hook entry integrity and path formatting. This access is necessary for the health check functionality and is confined to the local environment, with no network-capable tools permitted in the skill's allowed-tools configuration.
[INDIRECT_PROMPT_INJECTION]: The skill processes multiple repository files (rules, hooks, scripts, and READMEs) to perform consistency and integrity checks, which identifies a surface for indirect prompt injection. 1. Ingestion points: Files located in .claude/rules/, .claude/hooks/, .claude/scripts/, and .claude/skills/. 2. Boundary markers: Absent; the skill does not use specific delimiters or warnings to isolate ingested file content from the audit logic. 3. Capability inventory: The skill has access to Bash and Read tools. 4. Sanitization: Absent; the skill reads raw file contents to perform hashing and structural validation.

claude-health