spec-harness-build

Implement tasks from tasks.md with gated verification and learning loop routing.

28|2|Updated Oct 3, 2025
One-click install
npx skills add https://github.com/chohra-med/expo_boilerplate --skill spec-harness-build
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: spec-harness-build
Source: https://github.com/chohra-med/expo_boilerplate/tree/main/.claude/skills/spec-harness-build
Command: npx skills add https://github.com/chohra-med/expo_boilerplate --skill spec-harness-build

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill eliminates the risk of inconsistent, unverified feature implementation when using AI coding tools, by enforcing a structured spec-driven pipeline with strict gated verification to prevent hallucinated patterns, incomplete work, and rule drift.

Core Features & Use Cases

  • Gated Task Pipeline: Implements tasks from tasks.md one at a time, with separate implementer, tester, verifier, and reviewer agents to enforce quality at every step.
  • Ratchet Rule Enforcement: Ensures project rules only ever tighten over time, with all failures and corrections routed to a learning loop to eliminate recurring issues.
  • Use Case: Ideal for software engineering teams using AI coding assistants like Claude Code or Cursor to ship production features without pattern drift or missed verification steps.

Quick Start

Use the spec-harness-build skill to implement the first task listed in your project's tasks.md file, following the full spec-harness pipeline with gated verification and learning loop routing for any failures.

Frequently Asked Questions about spec-harness-build

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

FAQPage Schema
How do I prevent inconsistent implementations when using AI coding tools for feature development?

You can prevent inconsistent implementations by using a spec-driven gated pipeline that enforces separate implementer, tester, verifier, and reviewer steps for each task, ensuring no task is marked complete without passing verification.

How does gated verification work in spec-driven development?

Gated verification works by implementing tasks from a tasks.md file one at a time, passing each through distinct phases that test and verify against a goal.md file before review, routing all failures to a learning loop to eliminate recurring issues.

What is the ratchet rule in AI-assisted software engineering pipelines?

The ratchet rule ensures project rules only ever tighten over time during feature implementation, automatically routing all failures and corrections to a learning loop to prevent hallucinated patterns and eliminate rule drift.

Can I use this spec-driven pipeline with AI coding assistants like Claude Code or Cursor?

Yes, this pipeline is ideal for software engineering teams using AI coding assistants like Claude Code or Cursor to ship production features without pattern drift, enforcing append-only project rules and strict gated verification.

What file formats do I need to set up a spec-driven gated workflow?

You need to define tasks in a tasks.md file and outline objectives in a goal.md file to satisfy the structured, ratcheted development pipeline requirements for gated implementation and verification.

Why do my AI coding assistant implementations fail verification against project goals?

Implementations fail when bypassing separate verifier agents, but a structured spec-driven pipeline enforces strict gated verification against goal.md and routes failures to a learning loop, ensuring no incomplete work is marked complete.