ai-low-hanging-fruit

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

• Third-party content exposure detected (high risk: 0.70). In directory mode, the runtime reads outsider-authored free text from prd.md (and optionally tech-spec.md) located in the user’s Obsidian vault via Read, then uses that content to generate the LLM output—so any PRD/spec text not authored by the operating user can be injected.