The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute standard security analysis tools like slither, dylint, and golangci-lint to identify vulnerabilities. This is a legitimate and expected function for its stated purpose.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes external codebases and has access to command execution and file operations, which creates a potential surface for indirect prompt injection. (1) Ingestion points: User codebase files. (2) Boundary markers: None specified to separate code from instructions. (3) Capability inventory: Shell command execution and file system interaction. (4) Sanitization: No explicit sanitization or filtering logic is mentioned.

audit-prep-assistant