What problem does it solve? Applying an approved change often goes wrong when an agent edits before confirming scope and authority, forces TDD onto work that has no meaningful failing test, repeats the same failed fix, or reports unrun checks as passing. This Skill structures implementation into small, reviewable work units with explicit verification choices and honest completion reporting. ## Core Features & Use Cases - Verification-mode selection: Chooses TDD (Red-Green-Refactor) only when a meaningful failing test can express the expected behavior, and selects deterministic or inspection-based checks for documentation, static configuration, and behavior-preserving refactoring. - Scope and risk gates: Stops as Blocked before high-risk edits lacking authorized scope, controls, or residual-risk acceptance, and halts when a discovered shared cause exceeds the authorized boundary instead of adding workarounds. - Failure-loop control: Distinguishes repeated test observation from repeated failed implementation attempts and stops before a third equivalent edit without new evidence. - Use Case: After a design is approved to fix a config parser bug, use this Skill to write a failing focused test, make the minimal fix, run the regression suite, and report actual files changed, commands run, unverified scope, and residual risk. ## Quick Start Implement the approved bug fix in small units, confirm a failing test before editing, run the focused and regression checks, and report what remains unverified.
