The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data, including draft texts, sent messages, and user-provided reasons for changes. While this presents an attack surface where malicious instructions could be embedded in the text, the skill mitigates this by instructing the agent to pass user text via files (--reason-file) rather than shell interpolation. The processing logic is also largely contained within a Python script that outputs structured metrics and tags, reducing the agent's direct interaction with potentially malicious content.
[DYNAMIC_EXECUTION]: The scripts/delta.py utility employs dynamic module loading to find its dependency, stylometry.py. It searches a path provided by an environment variable (PERSONAL_TONE_STYLOMETRY) as well as relative paths from the script's location. This is a common pattern for internal skill dependencies but relies on the integrity of the local environment.
[COMMAND_EXECUTION]: The test suite scripts/test_delta.py utilizes subprocess.run to execute the main logic script during self-tests. This is an expected and legitimate use of command execution within a testing context.

voice-calibrate