review-to-release-workflow

Guide codebase review through four phases with approval gates and release verdicts.

16|Updated Apr 11, 2026
One-click install
npx skills add https://github.com/glaucia86/skills --skill review-to-release-workflow
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: review-to-release-workflow
Source: https://github.com/glaucia86/skills/tree/main/review-to-release-workflow
Command: npx skills add https://github.com/glaucia86/skills --skill review-to-release-workflow

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It prevents risky, unclear changes by forcing a structured codebase review, decision capture, approval-based implementation, and evidence-based release readiness.

Core Features & Use Cases

  • Phase 1 Discovery (codebase → QUESTIONS.md): performs a staff/principal-engineer style audit to identify ambiguities, risks, and missing decisions, producing a detailed QUESTIONS.md decision backlog.
  • Phase 2 Implementation (answered questions → approved plan → changes): reads the answered QUESTIONS.md, converts decisions into a scoped implementation plan, obtains explicit human approval, and applies only approved changes while generating IMPLEMENTATION_NOTES.md.
  • Phase 3 Verification (changes → evidence → VERIFICATION.md): verifies that approved decisions were implemented correctly and that scope boundaries were respected, producing VERIFICATION.md.
  • Phase 4 Release Readiness (artifacts → verdict → RELEASE_READINESS.md): assesses build, configuration, tests, deployment assumptions, operational readiness, and docs to produce a final Ready / Ready with caveats / Not ready verdict.

Quick Start

Tell the agent to run review-to-release-workflow for your repository and explicitly specify the release target (for example, production) before it begins discovery.

Frequently Asked Questions about review-to-release-workflow

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

FAQPage Schema
How do I enforce human-in-the-loop approval gates during a full codebase change?

Human-in-the-loop approval gates are enforced by structuring the workflow into four phases: discovery, implementation, verification, and release readiness. The agent stops at each gate, producing artifacts like QUESTIONS.md and VERIFICATION.md, until explicit human decisions are provided.

What is a go no-go release readiness assessment for codebase changes?

A go no-go release readiness assessment evaluates build, configuration, tests, deployment assumptions, and operational docs to produce a final verdict of Ready, Ready with caveats, or Not ready. It ensures risky changes are blocked before reaching production based on verified evidence.

How do I capture ambiguities and missing decisions before implementing codebase changes?

Capture ambiguities by running a discovery phase that audits the codebase like a staff engineer, identifying risks and unclear decisions. This process generates a QUESTIONS.md decision backlog that must be answered before any implementation plan is created.

How do I verify that approved implementation changes respect scope boundaries?

Verify scope boundaries during the verification phase by checking that approved decisions were implemented correctly. The agent produces a VERIFICATION.md artifact containing evidence that the changes match the approved plan without scope creep.

Can I use this release readiness workflow for small, low-risk codebase updates?

This workflow applies to full-codebase change requests where ambiguities, risk, or release impact require human-in-the-loop gates. Small, low-risk updates with clear scope may not need the four-phase process and explicit go/no-go decision artifacts.

What's the best way to structure a coding agent workflow for safe production releases?

Structure a safe production release workflow by enforcing phase ordering with stop conditions based on answered decisions and verification outcomes. Require the agent to produce and consume QUESTIONS.md, IMPLEMENTATION_NOTES.md, VERIFICATION.md, and RELEASE_READINESS.md across four distinct phases.