read-paper

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (medium risk: 0.30). The skill instructs removing author affiliations, acknowledgments, funding disclosures, and other provenance details—actions that can obscure conflicts of interest or provenance—constituting suspicious steering without clear malicious intent.

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

• Third-party content exposure detected (medium risk: 0.30). The /read-paper workflow ingests outsider-authored free text at runtime by resolving a user-supplied <url> (arXiv/DOI/PDF/Semantic Scholar/Open access/publisher abstract) and fetching full text or abstract via node ./scripts/resolve-url.js and node ./scripts/fetch-fulltext.js, then stripping noise and extracting six-paragraph signal content from that retrieved paper text.