truefoundry-docs

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly fetches and ingests remote, user-supplied URLs at runtime—see references/manifest-schema.md which states remote OpenAPI specs ("spec: type: remote, url: ...") are fetched and converted into MCP tools and that agent_card_url/hosted-a2a-agent sources are fetched at runtime—so untrusted third‑party content can change agent capabilities and thus enable indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly states that remote OpenAPI specs are "fetched at runtime and converted into MCP tools that control agent capabilities" (e.g. https://api.weather.example.com/openapi.json) and that hosted A2A agent cards are fetched at runtime and can influence agent behavior (e.g. https://research-agent.example.com/.well-known/agent.json), so these runtime-fetched URLs can directly control agent prompts/capabilities and are therefore risky.