Use for scanning the codebase and creating/updating the agent profile. Frequency: once initially, then only when agent code changes. Stale detection handles this automatically. Keywords: context, profile, scan, update profile, agent info Auto-activates on requests like: - "scan the agent", "analyze the agent" - "update profile", "refresh profile" - "what does this agent do?"
Use for creating test scenarios, contracts, and agent wrapper setup. Frequency: when test objectives change or new scenarios are needed. Reuses existing scenarios otherwise. Keywords: scenario, create scenario, init, contract, wrapper, agent setup Auto-activates on requests like: - "create a scenario", "init scenario" - "set up a test", "configure the wrapper"
Use for running tests against scenarios — includes data selection, bundle management, and test execution. Frequency: the most common entry point. Run repeatedly during the test-evaluate-fix loop. Keywords: test, run test, test my agent, simulation, run simulation, generate test data, synthesize inputs Auto-activates on requests like: - "run the test", "test my agent" - "run simulation", "execute test" - "generate test data", "synthesize inputs"
Use for prompt version comparison and stability testing. Frequency: when tuning prompts. Optional — use when A/B comparison is needed. Keywords: compare, comparison, prompt version, stability, A/B test, diff, stability test Auto-activates on requests like: - "compare prompts", "compare prompt versions" - "stability test", "run a stability test" - "v3 vs v4", "version comparison" - "run same input multiple times", "run the same input multiple times" - "compare prompts", "run comparison test"
Use for FluxLoop CLI installation, authentication, and project setup. Frequency: once per project. Automatically skipped if already set up. Keywords: setup, install, configure, login, auth, first time, get started Auto-activates on requests like: - "set up fluxloop", "install fluxloop" - "configure fluxloop", "log in to fluxloop" - "get started"
Use for evaluating test results, analyzing insights, and improving the agent. Frequency: after every test run. Core of the daily test → evaluate → fix loop. Keywords: evaluate, evaluation, improve, analyze results, insights, recommendations, re-test Auto-activates on requests like: - "evaluate the results", "analyze results" - "improve my agent", "what went wrong" - "generate insights", "review test results"