truefoundry-access-control

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

• Third-party content exposure detected (high risk: 0.90). This skill explicitly fetches and ingests external, potentially untrusted content at runtime—e.g., remote OpenAPI specs and agent_card_url/hosted-a2a-agent sources (references/manifest-schema.md), plus Git repo build sources and HuggingFace artifact downloads—which are converted into MCP tools or change agent behavior and thus could enable indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly allows registering MCP servers from a remote OpenAPI spec which is fetched at runtime and "converted into MCP tools that control agent capabilities" (example URL: https://api.weather.example.com/openapi.json), so this runtime-fetched URL can directly control agent behavior and is a high-confidence risk.