intelligems-knowledge

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md instructs running node scripts/search_knowledge.mjs ... which (per references/SCHEMA.md) loads references/sources.jsonl line-by-line and uses its public snippet/description text as evidence; those snippets are outsider-authored public web content, so the runtime LLM context can include outsider free text.