Plan

What This Skill Does

The Plan skill empowers OpenClaw to transition from reactive execution to proactive strategy. Instead of immediately diving into potentially erroneous actions, the agent evaluates the complexity, risk, and history of a request. It determines the appropriate 'Plan Depth' (L0 through L4) and generates a structured roadmap to ensure success. By tracking outcomes and specific strategy efficacy in outcomes.md and strategies.md, the skill continuously refines its decision-making process, learning when it is safe to auto-execute and when human validation is mandatory.

Installation

To integrate this skill into your environment, run the following command in your terminal: clawhub install openclaw/skills/skills/ivangdavila/plan Ensure you have the latest version of OpenClaw initialized to enable the tracking capabilities.

Use Cases

This skill is ideal for multi-step workflows that require precision. Use it for complex software refactoring, data migration tasks, or long-form research projects where success hinges on intermediate milestones. It is particularly valuable for developers who want to reduce the 'oops' factor in production environments, as the L4 planning level explicitly triggers human oversight for high-stakes, novel tasks.

Example Prompts

1. "Plan the refactoring of the user authentication module. Create an L3 plan focusing on backward compatibility and security."
2. "I need to migrate our customer database to the new schema. Generate an L4 plan with a clear rollback strategy and include a validation checkpoint for me to review."
3. "Evaluate the current task of generating a weekly market report. Should I auto-execute this or create a formal plan?"

Tips & Limitations

• Trust the Default: If you feel unsure about a task, default to planning. An extra five minutes of planning often prevents hours of debugging.
• Keep Documents Clean: Regularly review your outcomes.md and strategies.md files. They are the 'brain' of this skill; keeping them organized helps the agent learn faster.
• Evolution: Pay attention to the agent's prompts for 'Promotion'. If the agent suggests promoting a task to auto-execute, ensure you have reviewed at least five previous successful iterations before confirming.
• Context is Key: Always provide sufficient context in your initial prompt to help the agent correctly assess the risk and determine the appropriate planning depth.