Why is verification mandatory after subagent tasks?

Subagents may report completion while the actual filesystem remains unchanged. This skill ensures the main agent verifies the subagent's work before reporting back to the user.

What tools are allowed within the Proof of Work skill?

The skill utilizes a core set of tools including Write, Edit, Read, Bash, and Grep to gather evidence and perform validations.

What is the Proof of Work skill for Claude Code?

It is a specialized capability that forces the AI to provide concrete evidence, such as command outputs and git diffs, for any claim it makes about code changes or verification.

When should I apply the Proof of Work protocol?

It should be applied whenever making claims about test results, build statuses, linting, type checks, or any verifiable development assertion.

How does this skill prevent AI hallucinations?

By mandating a 'zero-tolerance' rule for trust, the skill requires the AI to use tools like git status and cat to prove that files were actually modified as reported.

Proof of Work

Name: Proof of Work
Author: TheBushidoCollective

byTheBushidoCollective

•

セキュリティとテスト

Mandates transparent and verifiable development workflows by requiring concrete evidence for every claim made about code quality, tests, and agent tasks.

概要

The Proof of Work skill transforms how Claude interacts with codebases by enforcing a 'show, don't tell' philosophy. It eliminates ambiguity and potential hallucinations by requiring the AI to provide concrete evidence—such as git status outputs, test logs, and file content spot-checks—for every modification or verification claim. This is particularly critical when delegating tasks to subagents, ensuring that work reported as 'complete' is actually reflected in the filesystem. By prioritizing transparency, this skill builds deep trust between the user and the AI, ensuring a professional and reliable software development lifecycle.

主な機能

Zero-tolerance policy for unverified claims about test or build status
Real-time validation of file modifications and creations
39 GitHub stars
Standardized reporting format for verification results
Automatic evidence gathering using Bash, Grep, and Read tools
Mandatory subagent verification via git and filesystem checks

ユースケース

Ensuring test suites have fully completed before reporting success
Verifying the actual completion of tasks delegated to subagents
Auditing codebase changes during complex refactoring sessions

概要

主な機能

Zero-tolerance policy for unverified claims about test or build status
Real-time validation of file modifications and creations
39 GitHub stars
Standardized reporting format for verification results
Automatic evidence gathering using Bash, Grep, and Read tools
Mandatory subagent verification via git and filesystem checks

ユースケース

Ensuring test suites have fully completed before reporting success
Verifying the actual completion of tasks delegated to subagents
Auditing codebase changes during complex refactoring sessions