How does this skill prevent Claude from skipping steps?

It uses auto-activation triggers and mandatory checkpoints that require Claude to verify the completion state of previous steps before proceeding to high-level actions like opening a Pull Request.

What happens if a mandatory step is missed?

The skill provides explicit blocking instructions and visual warnings at defined checkpoints (like Step 15 and 21), prompting the agent to return and complete the missing requirement.

Where is the workflow state stored?

State is tracked either through the built-in TodoWrite tool or a specialized workflow_state.yaml file located in the .claude/runtime directory.

Does this actually block Claude from making a PR?

In its current state, it acts as high-authority guidance and self-compliance logic. Hard-blocking requires integration with external pre-commit hooks or CI checks.

Can I customize the workflow steps?

Yes, the skill is designed to align with a project's DEFAULT_WORKFLOW.md file, which can be modified to suit your specific team requirements and mandatory gates.

Workflow Enforcement for Claude Code

Name: Workflow Enforcement for Claude Code
Author: rysweet

byrysweet

•

Productivity & Workflow

Enforces standardized agentic coding workflows through mandatory step validation and visual progress tracking.

This skill optimizes the reliability of AI-driven development by guiding Claude through a rigorous 22-step workflow designed for agentic coding. It mitigates 'completion bias'—the tendency for AI agents to skip mandatory code reviews or PR steps—by implementing persistent state tracking, visual progress indicators, and hard checkpoints. By defining mandatory gates at preparation, review, and feedback stages, it ensures that every task from initial workspace setup to the final merge follows a documented, repeatable process that maintains high code quality standards.

Key Features

01Integration with session-end verification to ensure full process compliance

02Auto-activation triggers for common development workflow start sequences

03Mandatory gate enforcement for critical steps like security reviews and PR feedback

0417 GitHub stars

05Visual progress bars and detailed status dashboards for real-time monitoring

06Persistent workflow state tracking via YAML or TodoWrite integration

Use Cases

01Ensuring AI agents don't skip mandatory security and code review steps before PR creation

02Tracking progress across complex, multi-session engineering tasks to prevent context decay

03Standardizing PR preparation and feedback implementation across distributed developer teams

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add rysweet/amplihack workflow-enforcement

For use in Claude.ai and ChatGPT

Download Skill