Why is this better than standard AI coding?

Standard AI may assume code works based on logic; this skill forces the AI to prove it works using real-world evidence from your terminal.

Can this skill be used for TDD?

Yes, it specifically includes patterns for Red-Green cycles, ensuring a regression test fails before a fix is applied and passes afterward.

How does this skill prevent false success claims?

It implements an 'Iron Law' that forbids any completion statement unless the AI has just run a verification command and confirmed the exit code and output.

Does it work with delegated agents?

Yes, it treats agent reports with skepticism and requires an independent check of the VCS diff and command output before accepting their work.

Rigorous Work Verification

Name: Rigorous Work Verification
Author: darthmolen

bydarthmolen

Security & Testing

Enforces evidence-based task completion by requiring fresh command output before making any success claims.

About

The Rigorous Work Verification skill instills a disciplined 'evidence before assertions' mindset in Claude's workflow. It prevents premature PRs and false success reports by mandating a strict gate function: every claim of completion must be preceded by fresh, successful execution of relevant verification commands like test suites, linters, or build scripts. By eliminating rationalizations and assumptions, this skill ensures that work is actually finished and functional before it is handed off, preserving trust and reducing development rework.

Key Features

Standardized patterns for TDD Red-Green verification and regression testing
Prevention of vague language and assumptions like 'should work' or 'probably fixed'
0 GitHub stars
Structured checklists for verifying requirements line-by-line before completion
Mandatory gate function requiring fresh command evidence for every success claim
Independent verification of agent-delegated tasks via VCS diffs and command checks

Use Cases

Ensuring third-party agent tasks are fully verified before being accepted into the main branch
Validating bug fixes with actual passing test evidence rather than just logic changes
Preventing broken code from reaching pull requests by mandating full build and test cycles

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add darthmolen/learning-knowledge-store verification-before-completion

For use in Claude.ai and ChatGPT

Download Skill

GitHub