The Agent Skills Directory

[Indirect Prompt Injection] (LOW): Potential attack surface for indirect prompt injection as the skill reads and processes the content of README.md files which may be influenced by external contributors.
Ingestion points: reads content from ideas/**/README.md and CLAUDE.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are defined in the skill logic.
Capability inventory: The skill possesses Edit, Read, Glob, and Grep tools, allowing it to modify files on the local filesystem.
Sanitization: No explicit sanitization or validation of the README content is performed before processing.
[Privilege Escalation] (SAFE): The skill does not request or use elevated privileges (e.g., sudo) and operates within a restricted directory scope (ideas/).
[Data Exfiltration] (SAFE): No network tools are included in the allowed-tools list, preventing data from being sent to external domains.

readmes