microsoft-foundry

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly instructs the agent to pull and read third‑party content (e.g., listing and downloading sample directories via the GitHub API in foundry-agent/create/create.md "GET https://api.github.com/repos/…/contents/…" and to use the WebSearchPreviewTool/Bing grounding tools that fetch live public web results), so the agent will ingest untrusted, user-generated web content and README/code files which can materially influence tool selection and subsequent actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The create workflow includes runtime download commands that fetch sample code from the Microsoft Foundry samples repo (e.g. https://api.github.com/repos/microsoft-foundry/foundry-samples/contents/... and https://github.com/microsoft-foundry/foundry-samples) via curl/gh and then runs the project entrypoint, so externally fetched content can execute code at runtime.