The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run and direct shell calls in SKILL.md to execute internal helper scripts (pick_exemplars.py, voice_check.py). These operations are limited to a private temporary directory created with mktemp -d to ensure process isolation and prevent data leakage into the user's project files. Findings from automated analysis related to these calls are considered benign within the context of the skill's intended workflow.
[DATA_EXPOSURE_EXFILTRATION]: While the skill reads a user's profile.json and past message samples, the workflow explicitly states that data is processed locally. External data transmission is scoped to the LLM inference provider for the purpose of style drafting, which is the primary purpose of the skill. No transmission to unknown or suspicious domains was found.
[REMOTE_CODE_EXECUTION]: The skill references a sibling dependency voice-profile/scripts/stylometry.py. The script pick_exemplars.py dynamically adds this path to sys.path to load the module. This is a local dependency resolution pattern within the skill's own package structure rather than remote code execution.

voice-write