cloudflare-sandbox
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of external data and code from web requests and remote repositories.\n
- Ingestion points: External data is accepted via
request.json()andsandbox.gitCheckout()within the documented worker patterns.\n - Boundary markers: The provided examples focus on SDK implementation and do not explicitly demonstrate prompt delimiters for isolating untrusted input within prompt contexts.\n
- Capability inventory: The skill makes use of
sandbox.exec(),sandbox.runCode(), andsandbox.writeFile()to perform operations inside isolated Linux containers.\n - Sanitization: The documentation includes a specific "Security Hardening" section that provides an input sanitization function and advises using the code interpreter API to manage risks.\n- [EXTERNAL_DOWNLOADS]: The skill references necessary official dependencies and container images.\n
- Evidence: The skill utilizes the
@cloudflare/sandboxnpm package and thecloudflare/sandboxDocker image, which are official resources from a well-known service provider.
Audit Metadata