The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local bash scripts (scan.sh, lint.sh, detect_dupes.sh) located in the ~/.claude/skills/skills-janitor/ directory. These scripts are intended to analyze the state of installed skills and plugins. While these are local resources associated with the plugin's vendor, execution of shell scripts involves inherent risks associated with command-line tools.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources (other skills and agents located in ~/.claude/skills and ~/.claude/agents) and possesses command execution capabilities (Bash). This creates a vulnerability surface where a malicious skill could attempt to influence the agent's behavior during the audit process.
Ingestion points: Files located in ~/.claude/skills, ~/.claude/agents, and ~/.claude/plugins/installed_plugins.json (as referenced in SKILL.md).
Boundary markers: None explicitly defined in the skill instructions to delimit content processed from the scanned skills.
Capability inventory: The skill uses the Bash tool to execute auditing scripts (referenced in SKILL.md).
Sanitization: No explicit sanitization or filtering of the scanned content is described in the prompt instructions.

janitor-report