drizzle

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly instructs the agent to spawn a research sub-agent that fetches and parses public Drizzle documentation (e.g., https://orm.drizzle.team/llms.txt and multiple /docs/* pages via WebFetch), so the agent ingests open third-party web content that can influence its actions.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly instructs the runtime sub-agent to fetch the documentation index (https://orm.drizzle.team/llms.txt) and to WebFetch documentation pages at runtime, and those fetched documents are then used to construct prompts/responses—so external content at https://orm.drizzle.team/llms.txt directly controls agent behavior.