eval-triggers
Warn
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute local Python scripts located in
evals/scripts/using shell commands within markdown code blocks. - [COMMAND_EXECUTION]: A potential command injection vulnerability exists in Step 3. The instruction
python3 evals/scripts/compare_trigger_runs.py --baseline <baseline-summary>.jsonsuggests that user-provided input (<baseline-summary>) is directly interpolated into a bash command. If an attacker provides input containing shell metacharacters (e.g.,;,&,|, or backticks), they could potentially execute arbitrary commands on the system.
Audit Metadata