cloudflare-worker-base
Pass
Audited by Gen Agent Trust Hub on Jul 7, 2026
Risk Level: SAFEPROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [PROMPT_INJECTION]: Metadata poisoning and inconsistencies were detected. The skill claims a future 'Last Updated' date of 2026-01-03 in its README and SKILL files, likely to present the content as more current than it is. Additionally, the author listed in the plugin configuration ('Jeremy Dawes') does not match the system-provided author context ('fiveonecode').
- [PROMPT_INJECTION]: The skill exhibits an Indirect Prompt Injection surface. 1. Ingestion point: The '/api/echo' route in 'assets/src/index.ts' reads the full request body using 'c.req.json()'. 2. Boundary markers: No delimiters or instructions are provided to distinguish user data from agent instructions. 3. Capability inventory: The worker possesses environment bindings and the ability to forward requests via 'ASSETS.fetch'. 4. Sanitization: No input validation or filtering is performed on the data before it is returned, which could allow untrusted content to influence the agent's context.
- [CREDENTIALS_UNSAFE]: The configuration file 'assets/wrangler.jsonc' includes a specific hardcoded Cloudflare 'account_id' ('0460574641fdbb98159c98ebf593e2bd') instead of a placeholder. Hardcoding specific identifiers in shared templates is a data exposure risk.
Audit Metadata