understand-knowledge

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md Phase 3/4 dispatches LLM “article-analyzer” subagents with each batch including content from scan-manifest.json (which is populated by parse-knowledge-base.py reading outsider-authored wiki .md files and index.md at runtime), so free text from those files enters the LLM context.