cloudflare-hyperdrive

Pass

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions for downloading the cloudflared binary from Cloudflare's official GitHub repository (github.com/cloudflare/cloudflared). This is a trusted resource from a well-known service for secure tunnels.\n- [INDIRECT_PROMPT_INJECTION]: The skill involves ingesting data from external databases, which is a known vector for indirect prompt injection.\n
  • Ingestion points: Database query results are retrieved and processed within Worker scripts (e.g., src/index.ts).\n
  • Boundary markers: No explicit delimiters or instruction-ignore warnings are included in the provided patterns to separate untrusted data from the agent context.\n
  • Capability inventory: The skill is capable of executing SQL queries and making network requests to database hosts.\n
  • Sanitization: The provided examples do not demonstrate explicit validation or sanitization of records fetched from the database.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 22, 2026, 04:58 PM
Security Audit — agent-trust-hub — cloudflare-hyperdrive