taxonomic-interest-classification

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

• Third-party content exposure detected (high risk: 0.70). The workflow’s Step 2 explicitly supports an “LLM-assisted” mapping approach that ingests the corpus’s free-text items (e.g., posts/comments/articles) into the agent’s LLM context; if that corpus includes outsider-authored text, it becomes indirect prompt-injection exposure.