research-analyst-system

Pass

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE]: No malicious patterns or security vulnerabilities were identified in the architectural descriptions or configuration files.- [NO_CODE]: The skill package contains only documentation and metadata; no executable scripts or logic are included.- [DATA_EXFILTRATION]: The skill describes saving reports to a local directory, which is a benign and expected behavior for a research tool.- [PROMPT_INJECTION]: Analysis of the indirect prompt injection surface (Category 8):
  • Ingestion points: The workflow specifies that researchers read project-related materials and use search tools like WebSearch and metaso (SKILL.md).
  • Boundary markers: There are no explicit instructions or delimiters provided to mitigate the risk of instructions embedded in external content.
  • Capability inventory: The system's capabilities are limited to reading data and writing structured reports to the local file system.
  • Sanitization: No methods for sanitizing or validating external input are defined in the provided files.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 14, 2026, 08:32 AM