The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: Downloads and executes the audit runner from the vendor's GitHub repository using uvx. The reference is pinned to a specific commit hash (07608b3b9e4e359bc0b93b3eca94bb78d67e369e) to ensure code integrity.
[COMMAND_EXECUTION]: Executes shell commands to perform the test suite audit, including running mutants, replaying tests, and checking removal candidates. These commands are triggered by the agent using defined bash snippets.
[EXTERNAL_DOWNLOADS]: Fetches the core agentic-verification tool from a remote Git repository at runtime during the audit process.
[INDIRECT_PROMPT_INJECTION]: The skill processes JSON and Markdown reports generated by the external runner. This represents an attack surface where malicious content in the audited project or the runner's output could attempt to influence the agent's interpretation.
Ingestion points: Reads report.json and report.md from the .test-suite-audit/ project directory.
Boundary markers: Instructions explicitly command the agent to interpret numbers only from the report and to avoid upgrading findings beyond what is documented.
Capability inventory: Shell command execution via uvx and file system read/write access limited to the project's subdirectories.
Sanitization: No specific validation or sanitization of the external report content is described.

agentic-test-audit