iterative-retrieval
Pass
Audited by Gen Agent Trust Hub on Aug 5, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill documents the 'Iterative Retrieval Pattern', a legitimate architectural approach for agents to progressively refine codebase context retrieval through a four-phase loop (Dispatch, Evaluate, Refine, Loop).
- [NO_CODE]: The skill consists entirely of markdown documentation and abstract pseudocode snippets. It does not include any executable scripts, configuration files, or shell commands.
- [SAFE]: The skill references an external guide on X (formerly Twitter), which is a well-known service, for additional context on subagent orchestration.
- [INDIRECT_PROMPT_INJECTION]: The skill documents a pattern for agents to ingest and process codebase data. This introduces a theoretical attack surface where retrieved files could contain malicious instructions designed to influence the agent.
- Ingestion points: Phase 1/Loop phase via 'retrieveFiles' functionality.
- Boundary markers: None specified in the pattern's instructional text.
- Capability inventory: The skill does not define or request specific agent capabilities like subprocess or network access.
- Sanitization: No sanitization or filtering of retrieved content is specified in the pattern.
Audit Metadata