paper-narrative

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

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md step 1 and kernel.py, the workflow ingests outsider-provided manuscript abstract/intro and figure captions (untrusted input) into paper_brief_prompt(abstract_text, figure_claims), then re-reads/edits the resulting brief and uses it to build narrative_review_task for the full deck.