twl:co-self-improve

Automating test scenario execution and retrospective workflows involves coordinating a controller with an observer to load scenarios, manage test projects, and summarize past observations. This enables deterministic execution and supports issue drafting from observation data.

Managing test projects and loading scenarios is handled by a controller that safely spawns an execution arm in a dedicated test-target worktree. This ensures reproducible results by maintaining deterministic test-project state throughout the scenario run.

Retrospective mode works by generating and summarizing past observation data collected during test execution. It loads this historical data to support issue drafting, allowing teams to analyze previous test runs and identify areas for project improvement.

Yes, you can run self-improvement workflows locally and manage real issues. The controller supports real-issues and local execution modes, enforcing mode detection from user input to safely spawn a co-autopilot within the isolated test-target worktree.

Yes, you need an observer tool acting as the execution arm to coordinate test management and scenario runs. The controller relies on this integration to automate workflows, load scenarios, and execute tests within the dedicated worktree environment.

Spawning a co-autopilot in a test-target worktree is limited to a dedicated, isolated environment to ensure safe execution. Mode detection must be enforced from user input, restricting the co-autopilot to deterministic scenario execution and test-project state management.