iterative-retrieval
Pass
Audited by Gen Agent Trust Hub on Mar 11, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted codebase data into the agent's reasoning loop, creating an attack surface for indirect prompt injection.
- Ingestion points: The 'DISPATCH' phase and the
retrieveFilesfunction described inSKILL.mdare points where external file content is introduced to the agent. - Boundary markers: The instructions do not specify the use of delimiters or 'ignore' directives to prevent the agent from executing instructions found within the retrieved files.
- Capability inventory: The pattern relies on the agent's ability to read local files, score their relevance, and identify gaps in context.
- Sanitization: There is no evidence of sanitization or filtering logic to detect or neutralize malicious content embedded in the codebase files before processing.
Audit Metadata