The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes command-line utilities such as ripgrep (rg), codedb, and colgrep to perform its primary function of auditing software systems. These tools are used for legitimate tasks like searching text, locating definitions, and summarizing code structure as part of the audit process.
[INDIRECT_PROMPT_INJECTION]: The skill acts as an audit engine that ingests untrusted external data, such as code repositories, third-party claims, or plan documents, to generate its reports. This represents a known attack surface where malicious content within the analyzed files could attempt to influence agent behavior.
Ingestion points: The skill ingests 'bounded targets' including repositories, diffs, data pipelines, and external documents as defined in SKILL.md.
Boundary markers: The instructions do not specify explicit delimiters or 'ignore' warnings to isolate ingested target data from the agent's core logic.
Capability inventory: The skill utilizes file-reading and search capabilities through tools like rg and codedb in packs/software-audit.md.
Sanitization: No explicit sanitization or filtering of external content is defined in the instructions.

conjecture-cascade