workflow-tools

What This Skill Does

The workflow-tools skill is a unified command-line utility designed to streamline complex AI agent workflows by consolidating four critical productivity modules: loop detection, parallel/serial decision-making, MCE (Minimal Context Engineering) file analysis, and subworkflow spawning. Rather than juggling multiple disparate skills, users can now handle technical friction—such as hanging tasks, bloated codebase files, or inefficient execution strategies—through a single, cohesive interface. By integrating with failure-memory and constraint-engine, this skill acts as a safeguard against incomplete work, scanning your workspace for unfinished markers like TODOs or placeholders, providing a 5-factor decision framework for parallelizing tasks, and automatically suggesting file splits when files exceed optimal complexity thresholds. It is built to keep your agentic cycles efficient, clean, and highly organized without requiring external API calls.

Installation

To install workflow-tools, ensure you have the necessary dependency context modules ready. Run the following in your OpenClaw terminal:

openclaw install leegitw/context-verifier
openclaw install leegitw/failure-memory
openclaw install leegitw/constraint-engine
openclaw install leegitw/workflow-tools

Once installed, the command /wt becomes available as your primary entry point for all sub-commands.

Use Cases

• Project Housekeeping: Before committing code, run /wt loops to ensure no deferred implementation details or forgotten placeholders remain in your current directory.
• Optimization Strategy: Use /wt parallel when dealing with multi-stage tasks to determine if your agent should process sub-tasks concurrently or sequentially, improving overall latency.
• Refactoring Support: When dealing with large legacy files, /wt mce identifies code blocks that exceed 200 lines, providing actionable suggestions to split them into more manageable, maintainable modules.
• Dynamic Scaling: Use /wt subworkflow to dynamically spawn child tasks for complex goals that require isolating specialized agent logic.

Example Prompts

1. "/wt loops ./src --exclude node_modules --pattern BUG,FIXME"
2. "/wt parallel --analyze --task='Refactor database schema and update API docs'"
3. "/wt mce ./services/auth_controller.py"

Tips & Limitations

• Precision: The loop detection works best when you keep consistent naming conventions for your markers (e.g., TODO, PLACEHOLDER).
• Scope: While the tool performs file analysis, it does not automatically edit files for you; it provides suggestions that you should review before applying. Always ensure your environment is backed up before acting on MCE split recommendations. The tool is local-only, ensuring your data never leaves your infrastructure.