faq
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell commands, specifically
git ls-filesandgrep, to identify the correct casing for documentation files and to search for information within the repository's source code. These operations are restricted to local repository analysis and do not incorporate user-supplied input in a manner that would allow for command injection. - [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection because it ingests untrusted data from user questions and repository source files (including comments and scripts) to generate FAQ entries. This is inherent to its primary function of summarizing repository information.
- Ingestion points: Untrusted data enters the agent's context through user-provided questions and the content of repository files processed during the research phase.
- Boundary markers: The instructions do not specify explicit delimiters or "ignore instructions" markers to isolate the content of scripts being researched from the agent's own task instructions.
- Capability inventory: The skill is restricted by its configuration to reading repository files and writing to a local FAQ document path. It is explicitly prevented from invoking other models and does not perform network operations.
- Sanitization: The skill instructions emphasize rewriting and distilling content for clarity and voice, though they do not mandate specific security sanitization of the researched data.
Audit Metadata