mastra

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow can fetch remote Mastra documentation from https://mastra.ai/llms.txt and arbitrary https://mastra.ai/.../*.md pages via the WebFetch tool, which ingests outsider-authored free text 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 (notably https://mastra.ai/llms.txt and related https://mastra.ai/*.md docs) and shows examples (WebFetch(...) and “fetch from https://mastra.ai/llms.txt”) that feed markdown into the agent context, so external content can directly control prompts and is relied on when embedded docs are absent.