iterative-retrieval
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a loop where an agent reads and evaluates project file content to refine its retrieval strategy, creating a vulnerability surface.
- Ingestion points: In Phase 2 (EVALUATE), the agent is instructed to read and process
file.contentfrom the repository. - Boundary markers: The instructions do not define or recommend using markers to delimit external data or instructions to ignore embedded commands within the files being retrieved.
- Capability inventory: The agent uses retrieved data to score relevance and dynamically generate new search keywords and focus areas, allowing the data to influence the agent's execution flow.
- Sanitization: The pattern lacks steps for sanitizing, escaping, or filtering the retrieved content before it is processed by the agent's reasoning logic.
- [EXTERNAL_DOWNLOADS]: The skill's 'Related' section provides a link to a guide hosted on X, a well-known social media service.
Audit Metadata