math-olympiad
Pass
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses local shell scripts (
scripts/check_latex.sh,scripts/compile_pdf.sh) to manage environment checks and LaTeX document compilation. - [COMMAND_EXECUTION]: Instructs the agent to generate and execute local Python and Bash scripts to perform mathematical verification and symbolic computations in 'Deep Mode'.
- [PROMPT_INJECTION]: Utilizes highly prescriptive prompts and adversarial subagent roles to ensure mathematical correctness; these are task-specific logic controls rather than safety bypass attempts.
- [DATA_EXFILTRATION]: Includes strong, explicit instructions forbidding network access and web searches during the solving process to maintain academic integrity and data security.
Audit Metadata