cloudflare-workers
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documentation includes patterns for handling external data from HTTP requests and WebSockets.
- Ingestion points: HTTP Request objects in the fetch handler and DurableObject fetch method, as well as WebSocket message events in the Room class (SKILL.md).
- Boundary markers: The examples do not explicitly show delimiters for isolating user data from instructions, which is standard for this type of runtime environment.
- Capability inventory: The code examples demonstrate capabilities for writing to databases (D1), key-value stores (KV), blob storage (R2), and triggering asynchronous tasks via Queues (SKILL.md).
- Sanitization: The skill correctly demonstrates sanitization for database operations by using prepare().bind() syntax to prevent SQL injection (SKILL.md).
- [SAFE]: The skill emphasizes secure development practices, specifically advising users to store sensitive credentials using wrangler secret rather than hardcoding them in configuration files or code.
Audit Metadata