mastra

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime workflow can fetch remote documentation from https://mastra.ai/llms.txt and other https://mastra.ai/.../*.md URLs (via the described WebFetch/remote-docs path), which is outsider-authored free text that may be injected into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs agents to fetch remote documentation at runtime (e.g., https://mastra.ai/llms.txt and agent-friendly .md pages like https://mastra.ai/reference/.../*.md via WebFetch) and to inject that markdown into the agent context to guide behavior, so the external URL is used at runtime and directly controls prompts/instructions.