Can the agent get stuck in an infinite loop?

No, the skill includes a safety mechanism that detects if an agent has already been blocked by a stop hook, allowing it to proceed on its next attempt to prevent infinite loops.

What is Deep Verification mode?

Deep Verification is an optional Layer 3 check that uses a sub-agent to read your task lists and plan files, ensuring that what the agent claims is finished actually matches the project state.

Does it slow down my development process?

The core checks are highly optimized, with regex running in under 1 second and intent analysis in 2-3 seconds. Only the optional Deep Verification layer takes longer (10-30s) for critical accuracy.

How does the Anti-Laziness Guard work?

It intercepts 'Stop' events in Claude Code and runs three parallel checks—regex matching, intent analysis, and optional deep verification—to ensure the agent hasn't skipped assigned tasks or used shortcuts.

How do I enable the deep check feature?

You can activate Layer 3 deep verification by creating a flag file with the command 'touch .claude/anti-laziness-deep-check' in your project directory.

Anti-Laziness Guard

Name: Anti-Laziness Guard
Author: swannysec

byswannysec

0•

生产力与工作流

Prevents Claude Code from skipping tasks or using shortcuts by implementing a multi-layered verification system during stop events.

The Anti-Laziness Guard is a comprehensive quality assurance plugin for Claude Code designed to combat 'context anxiety' and premature task termination. By utilizing a three-layer detection system—ranging from deterministic regex patterns to deep sub-agent verification—it ensures that agents do not skip assigned work, offer excuses for incomplete tasks, or use euphemistic language to justify shortcuts. This tool is essential for developers running long-running or critical workflows where absolute task completion and process adherence are mandatory, providing a robust safety net against AI rationalizations.

主要功能

01Parallel execution of stop hooks for minimal latency and maximum coverage during sessions

02Multi-layered detection including regex, LLM-based intent analysis, and deep file verification

03Built-in infinite loop prevention to ensure agents can gracefully exit after corrections

040 GitHub stars

05Tiered phrase taxonomy to block common AI shortcuts and rationalizations in real-time

06Deep verification mode that cross-references current tasks against plan files and TodoWrite entries

使用场景

01Executing complex refactoring tasks where every step of a migration plan must be completed

02Maintaining high documentation standards by preventing agents from summarizing instead of detailing

03Enforcing full code reviews and security audits without the agent skipping critical files

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add swannysec/robot-tools anti-laziness-guard

For use in Claude.ai and ChatGPT

Download Skill