How does this prevent 'hallucinations' of success?

It requires a fresh command run and explicit evidence (like exit codes or failure counts) before the agent is permitted to use completion language like 'Done' or 'Fixed'.

Why is the Verification Before Completion skill necessary?

AI agents may occasionally assume code works if it looks logically sound. This skill forces the agent to prove correctness through actual execution and output analysis.

Can this be used for documentation or non-code tasks?

Absolutely. It can be used to verify requirements via checklists or ensure that documentation builds successfully without broken links.

Does this skill work with any testing framework?

Yes, it is framework-agnostic. It relies on identifying the correct command for the specific environment and interpreting real-time terminal output.

Verification Before Completion

Name: Verification Before Completion
Author: dazeb

bydazeb

•

보안 및 테스팅

Enforces a mandatory verification protocol requiring empirical evidence before claiming task completion or success.

This skill implements a rigorous 'Gate Function' for AI-driven development, preventing premature or unverified success claims. It mandates that Claude must identify, execute, and analyze specific verification commands—such as unit tests, build scripts, or linters—before committing code or reporting a task as finished. By prioritizing evidence over assertions, this skill eliminates common failures like shipping broken functions, ignoring failed builds, or relying on 'should work' assumptions, ultimately building higher trust between the AI and the developer.

주요 기능

01Independent verification of delegated agent tasks and VCS diffs

02Evidence-based verification for tests, builds, and bug fixes

03Red-green cycle validation for robust regression testing

044 GitHub stars

05Mandatory gate function that blocks completion claims without fresh evidence

06Rationalization prevention to stop 'should work' or 'seems fine' responses

사용 사례

01Validating bug fixes against original symptoms before submitting a pull request

02Ensuring 100% test pass rates after complex refactoring sessions

03Verifying multi-step requirements through line-by-line evidence checklists

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add dazeb/claudeplaybook verification-before-completion

For use in Claude.ai and ChatGPT

Download Skill