kiro-impl

Automated TDD task execution dispatches per-task subagents in autonomous mode to implement approved tasks, running independent review cycles to enforce validation and prevent regressions. You provide a feature name to start the workflow.

Manual mode executes selected tasks by specific task numbers in the main context, bypassing per-task subagent dispatch. Both modes enforce the same validation loop, including design checks and reviewer verification.

The validation loop enforces alignment to design specs, performs selective commits, and runs reviewer verification on all implemented tasks. This formal process guarantees no regressions across both autonomous and manual execution modes.

Subagent dispatch is ideal for automating approved task implementation when you need independent review cycles and guardrails. It reduces manual toil by autonomously executing TDD workflows while maintaining strict regression prevention.

No specific dependencies or components are required to run autonomous TDD execution. You start by providing a feature name for autonomous mode, or optionally specify task numbers to run a subset in manual mode.

Without enforced review cycles, autonomous task execution risks unverified code changes and regressions. The formal validation loop aligns work to specs, controls commits, and verifies via reviewer checks to guarantee no regressions occur.