03-work

Execute plan implementation units with TDD enforcement and checkpoint-based resume.

29|2|Updated Apr 17, 2026
One-click install
npx skills add https://github.com/leing2021/super-pi --skill 03-work
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: 03-work
Source: https://github.com/leing2021/super-pi/tree/main/skills/03-work
Command: npx skills add https://github.com/leing2021/super-pi --skill 03-work

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

It prevents broken or unverified implementations by enforcing TDD gates while allowing you to resume execution precisely after interruptions, so your coding agent can reliably complete the implementation phase of a plan.

Core Features & Use Cases

  • TDD enforcement with hard gates: applies RED → minimal code → GREEN → refactor per unit and stops on any verification breach.
  • Checkpoint resume: tracks progress via session_checkpoint and skips completed units when re-run.
  • Plan-to-units execution: derives executable tasks from plan implementation units, including dependency analysis via task_splitter.
  • Controlled subagent execution: uses ce_parallel_subagent and ce_subagent only for bounded, non-interactive leaf tasks; otherwise runs inline and preserves stage integrity.
  • Evidence-first failure handling: triggers stop-the-line, preserves error evidence, diagnoses root cause with debug discipline, guards with regression tests, then resumes.

Quick Start

Use the skill when you already have a ready plan by telling your agent to load and execute your implementation units from a plan path, such as: execute 03-work with the plan file at docs/plans/plan.md as input.

Frequently Asked Questions about 03-work

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I enforce TDD gates during agent-driven coding workflows?

TDD gates are enforced by applying a RED to minimal code to GREEN to refactor cycle per implementation unit. The workflow stops immediately on any verification breach to prevent broken or unverified implementations.

Can I resume code execution from a checkpoint after an interruption?

Yes, you can resume code execution from a checkpoint. The workflow tracks progress using session checkpoints and automatically skips completed implementation units when re-run after an interruption.

How do I split plan implementation units into executable tasks for an AI agent?

Plan implementation units are split into executable tasks using a task splitter after analyzing dependencies. This ensures the agent receives bounded, non-interactive leaf tasks suitable for controlled subagent execution.

What is evidence-based debugging for agent execution failures?

Evidence-based debugging triggers a stop-the-line protocol upon failure, preserving error evidence. It diagnoses the root cause with debug discipline, guards the fix with regression tests, and then resumes execution.

Does this code workflow handle bare prompts or only structured plan paths?

This code workflow distinguishes between structured plan paths and bare prompts. It requires loading project workflow and testing rules to derive executable tasks properly from plan implementation units.

When should I transition to the review stage after TDD execution?

You should transition to the review stage after the TDD execution successfully completes all implementation units. The workflow explicitly recommends moving to the 04-review stage to finalize the build.