The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because its core workflow involves reading and performing qualitative analysis on untrusted skill files.\n
Ingestion points: The skill reads SKILL.md and referenced files from user-specified directories as part of both the static analysis script and the AI-driven judgment workflow.\n
Boundary markers: Absent. No instructions are provided to the agent to treat audited file content as data only or to ignore embedded instructions.\n
Capability inventory: The skill utilizes Read, Grep, and Bash tools (specifically for python, uv, rg, and find), which could be misused if the agent inadvertently executes instructions found in the audited content.\n
Sanitization: Absent. Raw text from audited files is passed directly into the agent's context for review.\n- [COMMAND_EXECUTION]: The skill executes a bundled Python script for static analysis using the Bash tool.\n
Evidence: The Workflow section of SKILL.md instructs the user to run python scripts/skill_audit.py <skill-path>.\n- [EXTERNAL_DOWNLOADS]: The skill recommends the installation of the pyyaml library.\n
Evidence: SKILL.md includes a recommendation to run pip install pyyaml. pyyaml is a well-known, widely used, and trusted library for parsing YAML data.

skill-audit