answering-research-questions
Pass
Audited by Gen Agent Trust Hub on Jul 6, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill identifies an attack surface for indirect prompt injection by ingesting and processing data from external scientific literature.
- Ingestion points: External research abstracts and full-text articles retrieved via
rp_searchandrp_fulltextin Phase 3 and Phase 4. - Boundary markers: The instructions do not specify the use of delimiters or provide the agent with guidance to ignore potential instructions embedded within the research text.
- Capability inventory: The agent is granted capabilities to write to the file system (
mkdir, file writing) and execute locally generated Python scripts. - Sanitization: No explicit sanitization or validation of the retrieved academic content is performed before the agent evaluates it for data extraction.
- [COMMAND_EXECUTION]: The workflow involves generating and executing Python scripts to handle research data.
- Evidence: Phase 7 and Phase 8 instruct the agent to create and run scripts such as
generate_summary.pyand JSON filtering snippets to consolidate findings and manage research sessions.
Audit Metadata