machine-purchase-roi-gate
Pass
Audited by Gen Agent Trust Hub on Aug 11, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructions explicitly prohibit the agent from accessing wallets, signing transactions, making payments, or broadcasting to any network, ensuring the 'ROI Gate' remains a read-only calculation tool.
- [EXTERNAL_DOWNLOADS]: The workflow references several external tool skills and dependencies, including
data-format-validationandcomplex-mathematics-engine. These are provided as links to the vendor's official pages on AgentPMT and ClawHub. - [COMMAND_EXECUTION]: The documentation provides installation commands using
npxandopenclawto add required dependencies from theAgentPMT/agent-skillsrepository. These are standard installation patterns for the specific agent platform and target the vendor's own official resources. - [DATA_FORMAT_VALIDATION]: The skill incorporates a specific validation step (Step 2) to check input JSON for malformed data, non-finite numbers, and logical inconsistencies (e.g., failure probability range) before passing data to the calculation engine, which acts as a security safeguard against indirect prompt injection or logic errors.
Audit Metadata