report
Pass
Audited by Gen Agent Trust Hub on Jul 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Uses vendor-specific CLI tools including
evo report,evo status,evo tree,evo frontier,evo show, andevo diffto retrieve and display recorded state from the local workspace for visualization and status checking. - [PROMPT_INJECTION]: The skill processes data from external run artifacts such as
outcome.jsonandbenchmark.logto summarize results. This represents an indirect prompt injection surface; however, the risk is minimal as the instructions strictly limit the agent to read-only reporting without any capability to execute content from these files. - [SAFE]: Implements robust security constraints by explicitly prohibiting the execution of benchmarks, Slurm commands (e.g.,
sbatch,srun), arbitrary scripts, or file modifications. This ensures the skill remains read-only and prevents unauthorized resource consumption.
Audit Metadata