literature-retrieval

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION] (LOW): Potential for indirect prompt injection through external literature data.
  • Ingestion points: Skill processes external search results (titles, abstracts) via literature-search and literature-filter.
  • Boundary markers: None explicitly defined in the workflow to separate untrusted data from instructions.
  • Capability inventory: Includes automated and agent-driven merging, filtering, and report generation.
  • Sanitization: No sanitization or escaping of external content is mentioned in the execution flow.
  • [DATA_EXFILTRATION] (SAFE): No hardcoded credentials, sensitive file access patterns, or unauthorized network requests were found.
  • [REMOTE_CODE_EXECUTION] (SAFE): No remote scripts are downloaded or executed. References to internal skills are structural and do not involve remote execution.
  • [COMMAND_EXECUTION] (SAFE): The provided Python snippets are illustrative pseudocode examples for logic flow and do not represent executable system commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:41 PM