environment-international

Pass

Audited by Gen Agent Trust Hub on Sep 24, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill's instructions are purely informational and procedural, focusing on scientific journal submission criteria and desk-reject heuristics.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided manuscript data to evaluate topic fit. While this constitutes a data ingestion surface, the skill lacks dangerous capabilities such as file system writing, command execution, or network exfiltration, rendering the risk negligible.
  • [COMMAND_EXECUTION]: No shell commands or script execution patterns were identified in the instructions.
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to search for official author guidelines on the Elsevier website and reference local resource files. These are expected behaviors for a research assistant skill and target trusted domains.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 24, 2026, 11:28 AM
Security Audit — agent-trust-hub — environment-international