The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill processes project files which could contain malicious instructions designed to hijack the agent's flow.\n
Ingestion points: Analyzes project files using Read, Grep, and Glob as listed in the allowed-tools section of SKILL.md.\n
Boundary markers: No specific instructions or delimiters are provided to the agent to ensure it ignores or treats as data any instructions found within source code files.\n
Capability inventory: The agent has access to Bash and Write tools, enabling potential file modification or command execution if triggered by an injection.\n
Sanitization: No sanitization or pre-processing of codebase content is mentioned.\n- Command Execution (LOW): The skill explicitly allows the use of the Bash tool to perform documentation tasks.\n
Evidence: The allowed-tools metadata field in SKILL.md includes Bash. While intended for running documentation generators like Sphinx or JSDoc, it represents a high-privilege capability that could be misused if the agent is compromised.

tech-writer