What is a Truth Score in this context?

A Truth Score is a reliability metric ranging from 0.0 to 1.0 that evaluates the correctness of code, agents, and tasks based on logic flow, security, and adherence to best practices.

Is the 0.95 accuracy threshold customizable?

Absolutely. You can set custom thresholds globally in the configuration or on a per-command basis to match the specific strictness required for your environment.

How does the automatic rollback feature work?

If a verification check falls below your defined threshold (defaulting to 0.95), the skill uses Git to instantly revert the codebase to its last known good state, preventing broken code from persisting.

Can I integrate this skill with GitHub Actions or GitLab CI?

Yes, the skill provides structured JSON output and CLI commands specifically designed to be used as quality gates within CI/CD pipelines.

What specific criteria are used for verification?

The system evaluates five core areas: code correctness (syntax/logic), security (vulnerabilities/secrets), best practices (SOLID/style), performance (complexity), and documentation accuracy.

AI Verification & Quality Assurance

Name: AI Verification & Quality Assurance
Author: Microck

byMicrock

•

セキュリティとテスト

Ensures high-quality agent outputs and codebase reliability through automated truth scoring, code verification, and fail-safe rollback systems.

概要

The Verification & Quality Assurance skill provides a robust framework for maintaining production-grade standards in AI-assisted development workflows. By implementing a sophisticated truth scoring system with a default 0.95 accuracy threshold, it evaluates code correctness, security, and best practices in real-time. This skill is essential for developers using Claude to automate complex tasks, offering peace of mind through automatic git-based rollbacks and comprehensive quality dashboards that track reliability trends across the entire software development lifecycle.

主な機能

Interactive web-based dashboards for live quality monitoring and trend analysis
81 GitHub stars
Instant automatic rollback for changes failing to meet the 0.95 accuracy threshold
Automated verification of code correctness, security, and best practices
Real-time Truth Scoring (0.0-1.0 scale) for code and agent reliability
Seamless CI/CD integration with structured JSON, CSV, and HTML reporting

ユースケース

Automating security audits and best-practice adherence for every code change
Monitoring the reliability of AI-generated code during large-scale refactors
Enforcing strict quality gates in automated CI/CD pipelines to prevent regressions

概要

主な機能

Interactive web-based dashboards for live quality monitoring and trend analysis
81 GitHub stars
Instant automatic rollback for changes failing to meet the 0.95 accuracy threshold
Automated verification of code correctness, security, and best practices
Real-time Truth Scoring (0.0-1.0 scale) for code and agent reliability
Seamless CI/CD integration with structured JSON, CSV, and HTML reporting

ユースケース

Automating security audits and best-practice adherence for every code change
Monitoring the reliability of AI-generated code during large-scale refactors
Enforcing strict quality gates in automated CI/CD pipelines to prevent regressions