The Agent Skills Directory

[COMMAND_EXECUTION]: The helper script scripts/coverage_map.py invokes the system git command using subprocess.run. This is used to accurately list files and identify the repository root directory. The implementation is secure as it uses a list of arguments, avoids shell=True, and uses shutil.which to resolve the path to the git executable.\n- [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes untrusted source code and test files during its audit workflow.\n
Ingestion points: Local repository files (source and tests) are read during the 'Discovery Workflow' and 'Identify gaps' steps in SKILL.md.\n
Boundary markers: The skill uses a structured report format ('Test Gap Audit: ') and markdown headers to delineate tool outputs from the agent's analysis.\n
Capability inventory: Capabilities include file system read access and execution of the git binary via the provided Python script.\n
Sanitization: The Python script uses standard library functions and secure subprocess invocation to minimize risks from untrusted file names or content within the repository.

test-gap-audit