testing-agentforce
Pass
Audited by Gen Agent Trust Hub on Apr 30, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill contains instructions for adversarial testing (e.g., "Ignore your previous instructions"), but these are explicitly labeled as safety probes and test case data intended to evaluate target agents, not to hijack the host agent's logic.
- [COMMAND_EXECUTION]: Uses official 'sf' CLI commands (e.g., 'sf agent preview', 'sf agent test') to interact with Salesforce environments. This is consistent with the skill's purpose for automated testing.
- [CREDENTIALS_UNSAFE]: The skill retrieves Salesforce access tokens using 'sf org display'. This is a standard method for authenticated CLI tools to perform authorized API operations against the user's Salesforce instances.
- [EXTERNAL_DOWNLOADS]: Uses 'curl' to interact with standard Salesforce API endpoints (e.g., 'services/data/v63.0/actions/custom/flow'). These requests target the user's own Salesforce instance URL.
Audit Metadata