The Agent Skills Directory

[DATA_EXPOSURE]: The skill utilizes file:read and file:write permissions to process user-provided documents and save a voice profile to workspace/soul/VOICE-PROFILE.md. These operations are local to the environment and consistent with the skill's declared functionality.
[PROMPT_INJECTION]: The skill processes untrusted external data (5,000+ words of writing) which presents a surface for indirect prompt injection.
Ingestion points: Reads text from files or user input during the "learn my style" command.
Boundary markers: None identified in the prompt templates.
Capability inventory: Limited to file:read and file:write within the workspace; no network or command execution capabilities detected.
Sanitization: No explicit sanitization of input text is mentioned.

style-clone