The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run commands to verify dynamic claims during the audit process (Step 4). This is a primary function of the tool and is constrained by instructions to use 'safe checks' and record the output for verification.
[DATA_EXPOSURE]: The skill is designed to read and inspect local project files, including source code, tests, and configuration files, to perform its audit. It generates artifacts and sidecars within a specific local directory structure (~/.agents/artifacts/).
[INDIRECT_PROMPT_INJECTION]: The skill possesses a broad ingestion surface as it is designed to analyze untrusted source code.
Ingestion points: Reads local project source code, callers, tests, and configuration files.
Boundary markers: None explicitly defined to ignore embedded instructions in the source code being audited.
Capability inventory: Includes command execution (to verify dynamic claims), file writing (artifacts), and file deletion (cleanup during close).
Sanitization: Instructions specify 'safe checks' but rely on the agent to distinguish between evidence and executable instructions within the audited files.
[SAFE]: No evidence of credential harvesting, unauthorized network exfiltration, or obfuscated content was found. The file deletion capability is part of a user-initiated 'Close' procedure for managing transient audit materials.

sus-audit