create-skill-autoresearch

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

• Third-party content exposure detected (medium risk: 0.30). The required Phase 3.4/4 evaluation runtime ingests user-supplied gold-standard/test-case text by reading TEST_CASE/INPUT (from input/), then sending it to an LLM as the user message while using the generated SKILL.md as system content.