rag-architect
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides tools (chunking_optimizer.py and retrieval_evaluator.py) designed to ingest and process text documents from user-specified local directories, creating a surface for indirect prompt injection.\n
- Ingestion points: Document content is read using standard file operations in
DocumentCorpus._load_documents(chunking_optimizer.py) andload_corpus(retrieval_evaluator.py).\n - Boundary markers: The scripts do not implement specific delimiters or safety instructions to the agent to disregard embedded prompts within the processed content.\n
- Capability inventory: The tools are restricted to text analysis, statistical calculations, and design logic; they do not contain capabilities for network access, subprocess execution, or dynamic code evaluation (eval/exec).\n
- Sanitization: No sanitization or filtering is applied to the document content before it is processed or displayed in analysis samples.
Audit Metadata