ai-integration
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches AI integration libraries including 'ai', '@ai-sdk/anthropic', '@ai-sdk/openai', and '@anthropic-ai/sdk' from official registries. These are established tools from well-known AI providers.
- [INDIRECT_PROMPT_INJECTION]: The skill provides a Retrieval Augmented Generation (RAG) implementation that ingests data from a database and interpolates it into prompts. Ingestion points: Untrusted database content enters the context via the 'ragChat' function in 'SKILL.md'. Boundary markers: The system prompt uses labels but lacks structural delimiters like XML tags or JSON blocks to isolate the context. Capability inventory: The skill performs database operations and calls external AI APIs. Sanitization: Although recommended in the included checklist, the code examples do not show explicit sanitization of the database-retrieved content.
Audit Metadata