cloudflare-workers-ai

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill implements patterns that ingest untrusted data from user requests and external URLs, which are then interpolated into AI prompts without robust sanitization in the provided templates.\n
  • Ingestion points: User-provided strings in templates/ai-text-generation.ts, templates/ai-embeddings-rag.ts, and external content fetched from URLs in templates/ai-vision-models.ts.\n
  • Boundary markers: The templates lack explicit delimiters or instructions to prevent the model from following embedded instructions, although references/best-practices.md provides conceptual sanitization examples.\n
  • Capability inventory: The skill facilitates AI inference (env.AI.run), vector database operations (env.VECTORIZE), cloud storage (env.BUCKET.put), and SQL database interaction (c.env.DB).\n
  • Sanitization: Validation and sanitization are documented as best practices but are not active in the core logic of the provided code templates.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 12:00 AM
Security Audit — agent-trust-hub — cloudflare-workers-ai