vibe-research

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md Core Loop step 2 explicitly requires the agent to "Search the web and gather current context" (including projects, communities, notable discussions and links) — untrusted public third-party content the agent must read and use to form briefs and drive next actions, enabling indirect prompt injection.