skool-rag

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection (Category 8) as it processes external data from Skool community posts and incorporates it into the LLM context.
  • Ingestion points: External content is retrieved from a Pinecone vector database in scripts/skool_rag_query.py.
  • Boundary markers: The script uses markdown headers (e.g., '### Source i') and horizontal rules ('---') to isolate retrieved content from the system instructions.
  • Capability inventory: The skill's query pipeline is limited to API calls to OpenAI, Pinecone, Cohere, and Anthropic; no shell execution or sensitive file system operations are triggered by the query logic.
  • Sanitization: Community content is interpolated directly into the generation prompt without secondary filtering for embedded instructions.
  • [EXTERNAL_DOWNLOADS]: The scripts facilitate network communication with official endpoints of well-known technology providers (OpenAI, Pinecone, Anthropic, Cohere) for standard RAG operations. These interactions are consistent with the skill's stated purpose and use official SDKs.
  • [CREDENTIALS_UNSAFE]: The skill properly manages credentials by utilizing environment variables (os.environ) rather than hardcoding sensitive keys. Placeholders provided in the documentation are for configuration guidance only.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 02:18 PM