check-open-access

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

• Third-party content exposure detected (high risk: 0.75). The skill resolves an identifier to a DOI and then performs an Unpaywall lookup at runtime (i.e., it fetches/reads third-party web/API content describing the article’s open-access locations), which is outsider-authored free text/metadata that can be ingested into the agent’s LLM context.