basin
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- Untrusted Data Ingestion Surface: The skill facilitates the ingestion of data from external HTTP sources and Cloudflare logs into R2 Iceberg tables. While this creates a potential surface for indirect prompt injection via the data pipeline, the skill provides specific guidance on implementing client-side validation using libraries like Zod to mitigate these risks.
- Controlled External Communication: The skill includes a feature to submit feedback to a public GitHub repository. This process is secured by a mandatory five-step user-approval workflow, which includes a review of the exact content being sent and explicit instructions for the agent to redact sensitive information such as API tokens, account identifiers, and private logs.
- Developer Tooling and Network Access: The skill utilizes standard developer tools like the Cloudflare Wrangler CLI and performs network requests to official Cloudflare and GitHub domains. These operations are essential to the skill's purpose and are performed using industry-standard patterns for authentication and secure transport.
Audit Metadata