What happens if an acceptance criterion is only partially met?

Issues are categorized by severity. If a core requirement is partially met or fails, it is marked as HIGH or CRITICAL severity, and the skill will recommend a REJECTED status, returning the task for remediation.

Can an AI agent approve its own code changes?

No. This skill enforces a 'Self-Approval Prohibition' policy. The agent that implemented the code cannot validate it; validation must come from a human user or a different designated reviewer.

Does this skill replace automated testing?

No, it complements it. Automated tests (Gate 3) and code reviews (Gate 4) must pass before this validation gate begins. This skill focuses on the final alignment between the code and the user's original requirements.

Why does the skill reject 'Looks good' as an approval?

To prevent ambiguity and accidental sign-offs, the skill requires the explicit keywords 'APPROVED' or 'REJECTED'. This ensures a clear, auditable trail of human intent.

Development Cycle Validation Gate

Name: Development Cycle Validation Gate
Author: LerianStudio

byLerianStudio

•

安全与测试

Enforces mandatory human approval and evidence-based validation of acceptance criteria before completing development tasks.

关于

The Development Cycle Validation Gate (Gate 5) is a critical quality assurance tool for AI-driven software engineering. It ensures that passing technical tests isn't the only metric for success by requiring explicit proof that every business requirement and acceptance criterion has been satisfied. The skill systematically gathers evidence, categorizes issues by severity, and enforces a strict 'no self-approval' policy, preventing AI agents from unilaterally completing tasks without verified human sign-off. This creates a reliable 'stop-and-verify' point that maintains high standards for production-grade code.

主要功能

Severity-rated issue categorization (Critical, High, Medium, Low)
17 GitHub stars
Evidence-based verification for every acceptance criterion
Prohibition of self-approval by the implementing agent
Comprehensive validation reporting with automated and manual evidence
Mandatory human-in-the-loop 'APPROVED' or 'REJECTED' workflow

使用场景

Implementing a standardized sign-off process in multi-agent AI development environments
Verifying that complex features meet specific business logic requirements before deployment
Auditing the gap between technical pass/fail tests and actual functional intent

关于

主要功能

Severity-rated issue categorization (Critical, High, Medium, Low)
17 GitHub stars
Evidence-based verification for every acceptance criterion
Prohibition of self-approval by the implementing agent
Comprehensive validation reporting with automated and manual evidence
Mandatory human-in-the-loop 'APPROVED' or 'REJECTED' workflow

使用场景

Implementing a standardized sign-off process in multi-agent AI development environments
Verifying that complex features meet specific business logic requirements before deployment
Auditing the gap between technical pass/fail tests and actual functional intent