research-brainstorm

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources, including literature search results and user-provided documents such as CVs or project directories. These external inputs represent a surface for indirect prompt injection where adversarial content could attempt to influence the agent's logic.
  • Ingestion points: The skill ingests untrusted data in Phase 2 (user documents/directories), Phase 4 (external search results from academic repositories), and Phase 8 (outputs from the find-data skill).
  • Boundary markers: The instructions do not define specific delimiters or "ignore instructions" warnings when processing these external inputs.
  • Capability inventory: The skill has the authority to write markdown files to the local working directory (Phase 9) and invoke the find-data skill (Phase 8).
  • Sanitization: There are no explicit instructions for sanitizing or escaping retrieved content before incorporating it into the final research brief or ongoing dialogue.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 11:01 AM
Security Audit — agent-trust-hub — research-brainstorm