The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute benchmark commands found in the project's own documentation (the 'Performance commands' block). This execution is the primary mechanism for generating the performance evidence required by the skill's checklist.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface where it ingests untrusted code diffs and project-specific guides to determine which commands to run. A malicious codebase could potentially provide a dangerous benchmark command. However, this risk is inherent to developer tools that execute scripts within a project context and is mitigated by the tool's specific scope and the agent's own execution guardrails.

review-perf