The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze and review user-provided code files (via $ARGUMENTS or manual selection), which serves as an ingestion point for untrusted data. If the analyzed code contains malicious instructions, the agent might follow them, though the skill's specific persona and structured output format provide some inherent constraints.
Ingestion points: The skill reads files passed via $ARGUMENTS or chosen by the user for design review as described in SKILL.md.
Boundary markers: There are no explicit instructions to the agent to ignore embedded instructions within the source code files being reviewed.
Capability inventory: The agent has the capability to read project files and is instructed to 'offer to fix the issues directly', which implies file-writing capabilities.
Sanitization: No specific sanitization or escaping of the content of the analyzed files is mentioned before the agent processes them.

rams