Can I skip verification for simple fixes?

No, this skill operates under a strict 'Low Freedom' policy, meaning every claim requires fresh evidence without exception to maintain absolute reliability.

Why is this skill necessary for AI coding?

It prevents the AI from making 'hallucinated' success claims by forcing it to run actual commands and provide proof that code works before moving to the next task.

How does it handle verbose test output?

It utilizes a test-runner agent that captures full logs in its own context and only returns a concise summary and specific failures to the main chat, preventing context pollution.

Does this work with task management systems?

Yes, it is designed to verify success criteria for 'beads' tasks and epics, ensuring every requirement is checked before a task is closed.

What happens if a verification command fails?

The AI is required to report the actual status with evidence, investigate the failure, and re-run verification until the success criteria are truly met.

Strict Task Verification

Name: Strict Task Verification
Author: withzombies

bywithzombies

•

Security & Testing

Enforces rigorous evidence-based verification before any task, bug fix, or build is claimed as complete.

This skill acts as a mandatory gate function that prevents premature completion claims by requiring fresh, documented evidence of success. It mandates running full test suites, build commands, and success criteria checks before any task can be marked as finished or a PR created. By utilizing a specialized test-runner agent to handle verbose output, it ensures that Claude maintains high standards of honesty and accuracy while keeping the conversation context clean and focused on results rather than logs.

Key Features

01Strict success criteria validation for project task management

02Integration with test-runner agents to prevent context pollution

03Evidence-based reporting for unit tests, builds, and bug fixes

04Mandatory verification gate for all status and completion claims

0558 GitHub stars

06Elimination of 'should work' or 'probably fixed' assumptions

Use Cases

01Confirming bug fixes by testing against original symptoms with fresh evidence

02Verifying complex task completion against specific success criteria in beads or task trackers

03Ensuring all tests pass before committing code or creating pull requests

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add withzombies/hyperpowers verification-before-completion

For use in Claude.ai and ChatGPT

Download Skill