The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for running benchmarks using standard repository tools, such as mise exec -- pnpm run test:bench. These commands are appropriate for the skill's stated purpose of performance testing.
[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to analyze existing source code and unit tests to design benchmarks, creating an attack surface where instructions embedded in project files could influence agent behavior.
Ingestion points: Component implementation files, unit tests, and existing benchmark files (SKILL.md).
Boundary markers: None explicitly mentioned in the instructions for separating source code data from the agent prompt.
Capability inventory: Execution of project build and test commands via pnpm and mise (SKILL.md).
Sanitization: No specific sanitization procedures are defined for the analyzed source code.

authoring-benchmarks