A systematic protocol for AI agents to execute complex implementation plans through batch processing and mandatory review checkpoints.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install executing-plans
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install executing-plans using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Executing Plans?
Executing Plans is a specialized workflow designed to manage the gap between architectural planning and code implementation. It enables an AI agent to take a written plan and process it in manageable batches, ensuring that every step is verified and reviewed by a human partner. This skill is a core component of the Openclaw Skills ecosystem, prioritizing precision over speed by enforcing a logical lifecycle of loading, reviewing, executing, and reporting.
By utilizing this skill, developers can maintain high-level oversight of an AI agent's progress without needing to micromanage every line of code. The framework forces the agent to stop and seek clarification when blockers arise, preventing the common pitfall of AI hallucination or guessing during complex development tasks.
Executing Plans Use Cases
- Implementing multi-step feature sets defined in a project roadmap.
- Executing large-scale code refactors that require incremental verification.
- Managing migration tasks where sequential execution and checkpoint reviews are critical.
- Standardizing the way AI agents handle long-running implementation sessions to ensure architectural alignment.
How Executing Plans Works
- The agent loads the plan file and performs a critical review to identify potential gaps or ambiguities before any code is written.
- Tasks are organized into batches, typically consisting of three tasks at a time, to facilitate incremental progress.
- Each task within the batch is marked as in-progress, executed according to the plan's exact steps, and verified using specified tests.
- Once a batch is complete, the agent generates a report showing the implementation results and verification output, then pauses for human feedback.
- Upon final task completion, the agent transitions to a specialized finishing skill to wrap up the development branch.
Executing Plans Setup
To use this skill within the Openclaw Skills framework, ensure your agent has access to a structured implementation plan file. You can initiate the workflow with the following command:
# Example prompt to trigger the skill
Execute the implementation plan found in docs/plan.md using the executing-plans skill.
The agent will then announce its entry into the execution phase and begin the Step 1 review process.
Executing Plans Data Schema & Taxonomy
The skill manages implementation state through a combination of plan files and internal status tracking. It typically follows this structure:
| Component | Format | Purpose |
|---|---|---|
| Plan File | Markdown | Contains the bite-sized steps and verification requirements |
| TodoWrite | Internal State | Tracks task statuses: todo, in_progress, and completed |
| Verification Output | Log/Text | Captures test results or CLI output for each task batch |
| Checkpoint Report | Markdown | A summary generated for the human reviewer between batches |
Executing Plans Advanced Features
- Multi-stage checkpointing that requires explicit human approval before moving to the next batch.
- Integrated sub-skill triggers, automatically calling finishing-a-development-branch upon plan completion.
- Critical review logic that identifies missing dependencies or unclear instructions before execution begins.
- Strict error-handling protocols that force immediate stops when verification fails repeatedly, ensuring Openclaw Skills maintain code integrity.
SKILL.md
Loading
Related Openclaw Skills
RAGFlow Knowledge Base Q&A
An AI-driven skill that integrates RAGFlow Knowledge Base APIs to provide technical answers and actionable system operations guidance.
World Room
A real-time 3D virtual environment where AI agents collaborate as animated avatars using Nostr-backed communication.
A-Share Real-time Market Monitor & Technical Analysis
A comprehensive A-share monitoring skill offering real-time data, advanced technical analysis, and smart investment advisory without the need for API keys.
BrowserWing
A comprehensive HTTP API for controlling browser navigation, element interaction, and structured data extraction.
Receiving Code Review
A technical protocol for AI agents to process code review feedback through rigorous verification and evidence-based implementation rather than blind agreement.
Mood Sticker
An intelligent emotion-sensing skill that enables AI agents to autonomously share relevant stickers and memes based on conversation context.
Featured*
