The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute performance benchmark commands (specifically a bench command) if defined in the project's 'agent guide'. This is an intended and documented capability for gathering performance metrics during the audit.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and acts upon data from the codebase being audited. Ingestion points: Audited repository diffs and the 'project agent guide'. Boundary markers: None defined to isolate instructions within the project guide. Capability inventory: Execution of local shell commands via the benchmark process. Sanitization: No explicit validation or filtering of the commands extracted from the project guide before execution.

review-perf