truefoundry-guardrails

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

• Third-party content exposure detected (high risk: 0.90). The skill's documentation explicitly states that remote OpenAPI specs (references/manifest-schema.md: "Remote specs are fetched at runtime and converted into MCP tools") and external agent_card_url/hosted-a2a-agent sources are fetched at runtime and that guardrail integrations with external endpoint_url fields route data to third-party services, meaning the agent will ingest untrusted third‑party content that can change tools/behavior.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly allows fetching remote OpenAPI specs at runtime (spec.url — e.g. "https://api.weather.example.com/openapi.json") which are converted into MCP tools that directly control agent capabilities, and also documents runtime-fetchable agent_card_url (e.g. "https://research-agent.example.com/.well-known/agent.json") and a Git repo to clone/build ("https://github.com/truefoundry/custom-guardrails-template") — all of which are fetched at runtime and can alter agent behavior or execute code, so they present a high-risk external dependency.