agents-sdk
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for building agents that ingest untrusted data from external sources.
- Ingestion points: Untrusted data enters the agent context through
routeAgentRequest, RPC methods decorated with@callable(), and incoming webhook payloads. - Boundary markers: The skill includes explicit instructions to validate inputs against strict schemas (Security Guardrail ASI01) and use formal schemas for inter-agent communication (Security Guardrail ASI07).
- Capability inventory: The agent has access to file system tools (
write_file,create_file,edit_file) and command execution tools (execute_command,wrangler_deploy). - Sanitization: The instructions mandate the use of parameterized queries (
this.sql.exec(query, ...params)) for all database operations to prevent injection attacks (Security Guardrail ASI05). - [EXTERNAL_DOWNLOADS]: The skill fetches documentation and technical references from Cloudflare's official repositories and developer portals.
- Source:
https://developers.cloudflare.com/agents/ - Source:
https://github.com/cloudflare/agents - Source:
https://developers.cloudflare.com/durable-objects/ - Purpose: These URLs are used to retrieve the latest SDK signatures, lifecycle documentation, and storage engine specifications.
Audit Metadata