scenario-inspiration

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

• Third-party content exposure detected (low risk: 0.10). The required runtime workflow does not read outsider-authored free text before selecting specific trusted items; it only performs controlled search calls over the team’s assets and the Scenario public catalog, while the open web lane is explicitly described as “not a search call” with no specified mechanism for ingesting arbitrary attacker text.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill instructs runtime fetching from open-web APIs (for example api.are.na/v2, collectionapi.metmuseum.org, api.openverse.org/v1, openaccess-api.clevelandart.org, api.artic.edu/api/v1/artworks/search) and explicitly reads returned metadata.prompt / downloaded assets to build prompts, so external content fetched at runtime can directly control the agent's prompts.