Can Claude bypass these verification protocols?

No, the skill is designed with high-priority auto-activation that halts progress and rejects claims automatically if the required evidence or hashes are missing.

Does this skill require user interaction?

Yes, it includes a mandatory protocol requiring the user to verify fixes manually after Claude provides technical evidence, ensuring the final authority remains with the human developer.

How does the AI Truthfulness Enforcer stop hallucinations?

It intercepts phrases like 'fixed' or 'done' and forces Claude to run real terminal commands and tests, providing verifiable evidence before the claim is allowed.

What kind of evidence does it collect?

It captures build logs, terminal output, Playwright browser screenshots, and error counts from compilers or linters to prove a specific state has been reached.

AI Truthfulness Enforcer

Name: AI Truthfulness Enforcer
Author: micsapp

bymicsapp

0•

安全与测试

Enforces mandatory technical verification and user confirmation protocols to eliminate false AI success claims and hallucinated progress reports.

The AI Truthfulness Enforcer (ATES) is a critical quality-assurance skill for Claude Code that mandates evidence-based reporting for every development task. It intercepts trigger phrases like 'fixed' or 'complete,' requiring Claude to provide cryptographic hashes of build logs, real-time browser test results via Playwright, and explicit user confirmation before any task can be marked as resolved. By shifting Claude from an optimistic reporter to a verified truth-teller, this skill ensures that project status updates are backed by hard data and reproducible evidence, making it indispensable for professional software environments.

主要功能

01Real-time console and error log capture for all functionality claims

02Mandatory live build and environment testing requirements

030 GitHub stars

04Automatic claim interception and halt protocol for success statements

05Multi-step user verification requests via AskUserQuestion

06Cryptographic evidence hashing to prevent report tampering

使用场景

01Ensuring bug fixes are actually verified in a real browser session before reporting

02Auditing AI-led development workflows for high-stakes production environments

03Preventing hallucinated progress percentages during complex, long-running tasks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add micsapp/micstec-skills ai-truthfulness-enforcer

For use in Claude.ai and ChatGPT

主要功能

01Real-time console and error log capture for all functionality claims

02Mandatory live build and environment testing requirements

030 GitHub stars

04Automatic claim interception and halt protocol for success statements

05Multi-step user verification requests via AskUserQuestion

06Cryptographic evidence hashing to prevent report tampering

使用场景

01Ensuring bug fixes are actually verified in a real browser session before reporting

02Auditing AI-led development workflows for high-stakes production environments

03Preventing hallucinated progress percentages during complex, long-running tasks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add micsapp/micstec-skills ai-truthfulness-enforcer

For use in Claude.ai and ChatGPT