rag-architect

Fail

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: CRITICALEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill links to an external documentation site (jeffallan.github.io) that has been flagged as malicious and blacklisted by automated reputation scanners.
  • [REMOTE_CODE_EXECUTION]: Automated security analysis has flagged the core instruction file (SKILL.md) as potentially malicious (FileRepMalware).
  • [DYNAMIC_EXECUTION]: The reference implementation for 'Late Chunking' in references/chunking-strategies.md uses the 'trust_remote_code=True' parameter when loading a model from Hugging Face. This setting allows for the execution of arbitrary Python code included in the model repository, representing a major security risk for untrusted models.
  • [INDIRECT_PROMPT_INJECTION]: The skill's primary function of ingesting and processing large amounts of external document data creates a significant attack surface for indirect prompt injection.
  • Ingestion points: Document loading and chunking logic described in SKILL.md and references/chunking-strategies.md.
  • Boundary markers: The skill does not implement or recommend robust boundary markers to isolate ingested document content from the agent's instructions.
  • Capability inventory: The skill possesses extensive capabilities for network communication via multiple API clients (OpenAI, Pinecone, Cohere, Qdrant) and local data management.
  • Sanitization: There is no evidence of text sanitization to filter out or neutralize instructions embedded in ingested data.
Recommendations
  • CRITICAL: 1 file(s) identified as malware by FileRep - DO NOT USE
  • AI detected serious security threats
  • Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Sep 5, 2026, 11:39 AM
Security Audit — agent-trust-hub — rag-architect