What is the 'Iron Law' of this skill?

The Iron Law states that no completion claims can be made without fresh verification evidence. 'I think it works' is never accepted; only fresh test or build logs count as evidence.

Does this work with my specific test suite?

Yes, the methodology is tool-agnostic. It instructs Claude to identify your specific test commands, run them, and verify the output against your project requirements.

How does this skill prevent AI hallucinations?

It requires a mandatory 'READ' step where the AI must capture and interpret actual terminal output, exit codes, and log details rather than assuming the code it wrote is correct.

When does the verification skill trigger?

It triggers on keywords like 'verify', 'is it done', and 'ready to merge', or via the verification_reminder hook after a task is completed.

Evidence-Based Verification

Name: Evidence-Based Verification
Author: TechDufus

byTechDufus

•

보안 및 테스팅

Enforces a rigorous methodology to ensure code changes are fully tested and verified before claiming completion.

The Verification skill implements the 'Iron Law' of software development for AI agents: no completion claims without fresh, verifiable evidence. It provides a structured gate function that guides Claude through identifying requirements, running fresh tests or builds, reading actual terminal outputs instead of assuming results, and confirming success against specific criteria. By eliminating assumptions and 'rationalizations,' it prevents premature pull requests and regressions, making it an essential tool for high-stakes coding tasks where reliability is paramount.

주요 기능

01Anti-Rationalization enforcement to prevent 'I think it works' assumptions

02Evidence-driven completion claims based on fresh terminal output

0385 GitHub stars

04Structured Gate Function for systematic Identify-Run-Read-Verify-Claim workflows

05Requirement-to-evidence cross-referencing for full feature coverage

06Integration with verification_reminder hooks for automated workflow triggers

사용 사례

01After complex refactoring to verify no regressions were introduced

02Before creating a Pull Request to ensure all local tests and builds pass

03When validating work performed by sub-agents before a final handoff

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add techdufus/oh-my-claude verification

For use in Claude.ai and ChatGPT

Download Skill