Micro-TDD Execution

SKILL: Micro-TDD Execution

Purpose

Execute autonomous, deterministic code modifications during hands-free (AFK) operation when macro-workflow is not explicitly active. This skill guarantees project hygiene, prevents principle drift, and strictly binds execution to automated verification while remaining extremely token-efficient.

1. Operating Context & Routing

• Trigger: Routed autonomously by Layer 2 Orchestration whenever a code change is needed for backend logic, database tables, custom hooks, utilities, or state managers.
• Precedence: Acts as a Layer 3 Autonomous Skill. It is strictly subordinate to Core Rules (.agents/rules/) and explicit User Requests, but enhances default high-level operational guidelines in AGENT.md.
• Communication Protocol: Enforces Silent Execution Mode. Natively suppress inner monologue, execution step narration, and line-by-line code explanations.

---

2. Execution Paths

Fast-Track A: Silent Logic Cycle (Default)

Apply strictly to all pure logic, state updates, API mutations, hooks, and backend functions. Do not write production code until a failing test exists.

1. Isolate & Specify (RED):
   • Write exactly one minimal, target-focused unit test asserting the exact change or new capability.
   • Run the test suite natively (npm test, vitest, pytest, etc.).
   • Validate Red: Confirm the test fails specifically due to the absence of functionality—not due to runtime compile errors or typos.
2. Implement & Pass (GREEN):
   • Write the absolute simplest, non-speculative production code required to satisfy the failing test.
   • Re-run the test suite.
   • Validate Green: Confirm the new test passes and no regression errors exist across the remaining suite.
3. Clean Diffs (REFACTOR):
   • Clean up code formatting using your active code simplifier protocols.
   • Output only the passing test suite results and a 1-line summary: [DONE] [[ID]] verified.

Fast-Track B: Visual & Layout Bypass

Apply strictly to pure CSS/Tailwind, static asset placements, or presentation markup adjustments where automated logic assertions are fragile.

1. Structural Shield Audit: Verify the target component is not listed under IMMORTAL_COMPONENTS inside DESIGN.md. If shielded, halt execution immediately and notify Layer 2.
2. Environment Execution: Boot the local development or preview server environment.
3. Visual Verification Matrix: Complete a strict manual verification checklist:
   • Validate design token alignment (e.g., oklch, padding spacing scales) against layout rules.
   • Confirm cross-viewport scaling (mobile/desktop responsive breakpoints).
4. Log Output: Emit a concise layout compliance log specifying the layout changes verified.

---

3. Mandatory Safety Guardrails

The "Stuck" Protocol (Design Alert)

• Trigger: If the unit test setup or mock requirement exceeds 50 lines of code, demands mocking more than 3 system dependencies, or generates circular failing loops.
• Action: Freeze execution immediately. Exit Layer 3 execution. Elevate control back to the Layer 2 Orchestrator to execute the constitutional ASK protocol. Present exactly two architectural options to simplify the application design or API surface area to the user.

The Anti-Regression Bug Loop

• Trigger: A defect, bug, or incorrect execution behavior is caught or reported.
• Action: Before editing any production or application files, write a targeted regression test that explicitly reproduces the reported failure state. Verify the test is RED on the current codebase, then route back through Fast-Track A to eliminate the bug.