cloudflare-workflows
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill templates define entrypoints that ingest external data from workflow events, which is an expected surface for indirect prompt injection in automated workflows.\n
- Ingestion points: Workflow entrypoints in
templates/basic-workflow.ts,templates/scheduled-workflow.ts, andtemplates/workflow-with-events.tsingest external data via theevent.payloadobject.\n - Boundary markers: The provided templates do not implement specific boundary markers or instructions to delimit untrusted data from processing logic.\n
- Capability inventory: The skill facilitates network operations via
fetch()and persistent storage access viaD1DatabaseandKVNamespacebindings.\n - Sanitization: Database interactions correctly utilize parameterized bindings (e.g.,
prepare().bind()). However, some templates interpolate variables directly into URL strings (e.g.,fetch(https://api.example.com/users/${userId})), leaving sanitization responsibilities to the end-user.\n- [EXTERNAL_DOWNLOADS]: The documentation and setup guides reference the installation and use of official Cloudflare development tools and libraries.\n - Evidence:
SKILL.mdcontains instructions fornpm create cloudflare@latestand the installation of@cloudflare/vitest-pool-workers. These resources belong to a well-known service provider and are documented neutrally.
Audit Metadata