deep-research

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

• Third-party content exposure detected (high risk: 0.85). In full/lit-review/systematic-review modes, the pipeline’s Phase 2 bibliography_agent performs runtime external web/database discovery and ingests scraped/returned bibliographic text (e.g., titles/abstracts) from outsider-authored sources into the agent context for downstream synthesis/reporting.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)