Testing Agentforce

This handles the full testing lifecycle for Agentforce agents, from smoke tests during development to regression suites in CI/CD. You get two modes: ad-hoc preview testing with local traces for quick iteration, and persistent test suites deployed to the org for batch runs. The real strength is the fix loop, it auto-derives test cases from your agent's subagents and actions, runs them, analyzes the trace JSON to diagnose failures like topic misrouting or ungrounded responses, then suggests targeted fixes to your YAML. It also forces a safety verdict after running adversarial probes, which is smart because it's easy to skip that step. Works directly with sf CLI commands, no separate scripts to manage.