How does this skill help prevent bugs in AI-generated code?

It replaces 'hope-based' development with a mandatory verification phase that compares the final state of the code against the original requirements and runs automated quality checks.

What are the different verdict levels in the Verify Report?

The skill renders four verdicts: STRONG (all met), SOUND (minor issues only), UNSTABLE (important issues present), and FAILED (critical requirements missing).

Can this skill be used for automated code reviews?

Yes, it is specifically designed to act as a rigorous orchestration layer that audits work, identifies gaps, and categorizes issue severity.

What is the Reflexion methodology used in this skill?

It is a verification technique where the AI agent systematically gathers context and reflects on its own work to identify errors and iterate until specific success criteria are met.

Reflective Verification

Name: Reflective Verification
Author: twofoldtech-dakota

bytwofoldtech-dakota

セキュリティとテスト

Implements a systematic reflection methodology to verify task completion and ensure engineering precision through rigorous quality assurance.

概要

The Reflective Verification skill provides a structured framework for quality assurance based on the 'Reflexion' research paper. It solves the common problem of AI agents declaring success prematurely by mandating a systematic loop of gathering context, comparing work against requirements, and identifying gaps. By classifying issues by severity and rendering a final verdict (Strong, Sound, Unstable, or Failed), this skill ensures that every piece of code or documentation meets high-precision 'Industrial Clarity' standards before the task is considered finished.

主な機能

Automated integration with test suites, linters, and type checkers
Requirement-to-achievement gap analysis for precise goal tracking
0 GitHub stars
Systematic 5-phase Reflection process (Gather, Compare, Identify, Classify, Decide)
Severity-based issue classification for critical, important, and minor bugs
Standardized Verify Report generation for transparent engineering logs

ユースケース

Systematically auditing AI-generated pull requests before human review
Verifying documentation accuracy against actual codebase implementations
Finalizing complex API features to ensure all edge cases and validations are met

概要

主な機能

Automated integration with test suites, linters, and type checkers
Requirement-to-achievement gap analysis for precise goal tracking
0 GitHub stars
Systematic 5-phase Reflection process (Gather, Compare, Identify, Classify, Decide)
Severity-based issue classification for critical, important, and minor bugs
Standardized Verify Report generation for transparent engineering logs

ユースケース

Systematically auditing AI-generated pull requests before human review
Verifying documentation accuracy against actual codebase implementations
Finalizing complex API features to ensure all edge cases and validations are met