The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes Python scripts used for testing and mutation that execute other local scripts within the skill package using the subprocess module.
Evidence: scripts/mutate_prove_ac.py and scripts/test_prove_ac.py call subprocess.run to invoke the core prove_ac.py logic using the current Python executable.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes Acceptance Criteria (AC) from potentially untrusted external documentation and repository source code to produce a final assessment for the agent.
Ingestion points: requirements.md, synthesis files provided via the --ac-file argument, and various source files within the target repository scanned by scripts/prove_ac.py.
Boundary markers: The skill instructions include a logic gate requiring a Code Review: Approved status from a separate reviewer skill before proceeding.
Capability inventory: The skill performs extensive file system traversal and reading using os.walk and path.read_text.
Sanitization: The prove_ac.py script extracts tokens using regular expressions but does not perform sanitization or filtering of the extracted content before it is processed by the agent's residual reasoning loop.

ns-judge