The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads local project files to match "house style" and writes new documentation files. These are expected behaviors for a documentation tool. No sensitive file paths (e.g., .ssh, .aws) or network exfiltration patterns were detected.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it processes untrusted data from conversation history and existing project files. However, it lacks high-risk capabilities like network access or shell command execution, limiting the potential impact of such an attack.
Ingestion points: Reads conversation history and neighboring documentation files (SKILL.md).
Boundary markers: None explicitly defined in instructions.
Capability inventory: Local file system read and write.
Sanitization: None specified for the processed content.

to-ctx