prior-art-search

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

• Third-party content exposure detected (high risk: 0.80). This skill explicitly queries and retrieves public third-party patent content (Google BigQuery Patents public dataset and USPTO API / internet access as shown in SKILL.md and prior_art_types.py) and ingests abstracts, claims, and full patent text for analysis, so the agent will read and interpret external untrusted web-hosted documents that could contain indirect prompt-injection content.