alibabacloud-terraform-codegen

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.78). Yes—Step 4.2 performs WebFetch of provider documentation from GitHub raw URLs (outsider-authored public web content), and Step 4.3 then recites that fetched free text into the agent’s LLM context.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly WebFetches provider docs at runtime (e.g. https://raw.githubusercontent.com/aliyun/terraform-provider-alicloud/master/website/docs/{r|d}/<doc_name>.html.markdown and the GitHub blob doc URLs listed in the local catalog) and uses that fetched content to construct the recitations and generated HCL, so the remote URL(s) directly control agent behavior.