The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external codebase content, creating a potential surface for indirect prompt injection if analyzed files contain malicious instructions.
Ingestion points: Uses Read, Grep, and Glob to read and search source code files in SKILL.md.
Boundary markers: The skill lacks explicit instructions to treat code content as data only or to ignore embedded instructions.
Capability inventory: Access is restricted to file reading and searching; no capabilities for network operations, file writing, or shell execution are present.
Sanitization: The skill performs no sanitization or filtering of the ingested code content.

architecture-review