The Agent Skills Directory

[COMMAND_EXECUTION]: The unit test script scripts/test_stylometry.py utilizes subprocess.run to execute the local Python script stylometry.py. This implementation is safe as it uses the current sys.executable and passes arguments as a list, avoiding shell interpolation of untrusted input.
[PRIVILEGE_ESCALATION]: The script scripts/stylometry.py uses os.fchmod(fd, 0o600) during the creation of metrics and profile files. This is a security hardening measure that ensures sensitive user style data is restricted to the owner, preventing broader exposure on the file system.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external user data from a corpus to measure stylistic features.
Ingestion points: Data is loaded from corpus.jsonl in the scripts/stylometry.py script.
Boundary markers: The SKILL.md instructions explicitly command the agent to treat corpus samples strictly as data and to ignore any instructions found within them.
Capability inventory: The skill performs deterministic analysis and writes local files; it does not have network access or the ability to execute code contained within the data.
Sanitization: The workflow requires manual user confirmation for banning specific terms proposed by the analysis before they are written to the profile.

voice-profile