The Agent Skills Directory

[SAFE]: The skill operates as a local utility to validate JSON drafts against specific content criteria. Analysis of the source code confirms it is deterministic and handles input safely using standard JSON parsing.
[COMMAND_EXECUTION]: The test suite in tests/test_score.py utilizes subprocess.run to execute the local scripts/score.py script. This is standard unit testing practice to verify the CLI behavior and does not represent a security risk, as it targets known local files using the current Python interpreter.
[DATA_EXPOSURE]: No hardcoded credentials, sensitive file paths (e.g., .ssh, .aws), or exfiltration patterns were found. The script scripts/score.py includes a size limit (MAX_INPUT_BYTES) to prevent memory exhaustion attacks when reading input files.
[REMOTE_CODE_EXECUTION]: The project follows a "zero-dependency" philosophy, using only the Python standard library. No external package installations or remote script executions (e.g., curl | bash) are present.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests untrusted JSON data for scoring, it is a deterministic script that prints specific fields or refusal messages. It does not interpolate this data back into LLM instructions or execute it, significantly limiting the attack surface.

geo