brand-docx

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime LLM context is populated by the model-driven comprehend step via python scripts/cli.py comprehend-input --name <brand>, which prints a bounded {facts, excerpt} bundle that includes excerpt text extracted from the outsider-authored template .docx (user-provided file), so free-form template text can enter the agent’s LLM context through that path.