rag-architect
Pass
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were detected in the skill instructions or provided code examples.
- [CREDENTIALS_SAFE]: Code examples correctly use placeholders like 'YOUR_API_KEY' for service authentication, preventing credential exposure.
- [EXTERNAL_DOWNLOADS]: The skill references standard machine learning and data processing libraries available on official package registries (PyPI).
- [DATA_EXPOSURE_AND_EXFILTRATION]: While the skill design involves processing external documents and queries (standard for RAG), it does not contain logic to access sensitive system files or exfiltrate data to unauthorized domains.
- [INDIRECT_PROMPT_INJECTION]: As a RAG architecture skill, it inherently describes a system that ingests untrusted data (documents and user queries).
- Ingestion points: Document content in the chunking stage (
raw_docs) and user queries in the retrieval stage (query). - Capability inventory: The code examples perform API calls for embeddings and reranking, and interact with a local vector database instance. No dangerous system execution capabilities are present.
- Boundary markers/Sanitization: The implementation examples are high-level blueprints; developers implementing these designs should apply standard LLM security practices for the final integrated system.
Audit Metadata