The Agent Skills Directory

[Prompt Injection] (LOW): The skill is vulnerable to indirect prompt injection as it ingests and processes untrusted data from the user's workspace without explicit isolation.
Ingestion points: Untrusted source code and project documentation are ingested via Read, Grep, and Glob tools (referenced in SKILL.md).
Boundary markers: Absent. There are no instructions defining delimiters or clear boundaries to distinguish between the agent's instructions and the content of the files being documented.
Capability inventory: The agent has access to Bash, Write, and Edit tools, which could be leveraged if the agent inadvertently follows instructions embedded within the analyzed code.
Sanitization: No sanitization or filtering of ingested file content is specified.

documentation