WorldThreatModel

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

• Third-party content exposure detected (medium risk: 0.30). In TestIdea the workflow reads ~/.claude/LIFEOS/MEMORY/RESEARCH/WorldModels/{horizon}.md files at runtime and ingests the outsider-authored “idea/strategy/investment” from the user prompt (Step 2) to run analysis across horizons (Step 3), so free text provided by a non-trusted user is directly processed alongside trusted model text.