The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection. It is designed to ingest and analyze external codebases using tools like Read, Grep, and Glob. Maliciously crafted code files could contain hidden instructions intended to influence the agent's behavior during the optimization process. The procedure does not specify the use of strict boundary markers or content sanitization when interpolating code content into the analysis flow.
[COMMAND_EXECUTION]: The skill requires the Bash tool to perform performance profiling and benchmarking in Step 1 and Step 3. This necessitates the execution of the code being analyzed. If an attacker provides a malicious codebase for optimization, the benchmarking process could trigger the execution of arbitrary commands within the agent's runtime environment. This risk is tied to the primary purpose of the skill as a developer tool.

chrysopoeia