fabric-builder

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md instructs using scripts/fabric.py to fetch module README/variables/outputs/schemas from the public GitHub repo at runtime; scripts/fabric.py downloads those files as UTF-8 text and prints them to stdout (e.g., README/TF/schema content), which the agent would then ingest into the LLM context—an outsider-authored free-text source from public web content.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill requires at-runtime fetching of module README/variables/outputs from GitHub (e.g., https://raw.githubusercontent.com/GoogleCloudPlatform/cloud-foundation-fabric/master and the GitHub API https://api.github.com/repos/GoogleCloudPlatform/cloud-foundation-fabric) via scripts/fabric.py, and that fetched content is used to control how the agent generates Terraform module calls, so this is a runtime external dependency that directly controls agent prompts/instructions.