An advanced planning and tracking skill that decomposes complex requests into manageable, dependency-aware subtasks with a built-in learning loop.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install task-decomp
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 task-decomp 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 Task Decomposition?
Task Decomposition is a sophisticated framework designed for AI agents to handle intricate, multi-stage projects. By breaking down high-level requests into granular, sized tasks, it enables Openclaw Skills to manage dependencies, parallelize workflows, and track progress with high precision. This skill transforms a simple agent into a project manager capable of navigating technical debt, complex builds, and multi-tool integrations.
At its core, the skill focuses on persistence and evolution. By storing plans as markdown files within the workspace, the agent maintains a clear state that survives restarts. Furthermore, it incorporates a learning loop that analyzes completed plans to improve future sizing and dependency mapping, ensuring the agent gets smarter with every project it completes.
Task Decomposition Use Cases
- Requests involving three or more steps with logical interdependencies.
- Projects requiring coordination across multiple specialized tools or sub-agents.
- Multi-phase software builds, such as API deployments or database migrations.
- Situations where the user requests a breakdown, plan, or strategic approach to a problem.
- Long-running workflows that need to be tracked and resumed across different sessions.
How Task Decomposition Works
- Analyze the full user request to identify boundaries between tools, contexts, and outputs.
- Decompose the request into specific tasks categorized by size (S, M, or L) to ensure honest progress reporting.
- Create a structured plan file in the plans/ directory, noting explicit dependencies and opportunities for parallel execution.
- Execute tasks in the defined dependency order, updating the markdown plan immediately after each task completion.
- Adjust the plan dynamically if context changes or failures occur, using a revision counter to track iterations.
- Perform a post-completion retrospective to extract successful patterns and update the global patterns library.
Task Decomposition Setup
To enable this skill within your workspace, ensure the agent has permission to manage a dedicated plans directory. You can initialize the environment using the following commands:
mkdir -p plans/
touch plans/patterns.md
Once the directory is ready, the agent will automatically begin using this directory to store active, archived, and pattern-learning files associated with Openclaw Skills.
Task Decomposition Data Schema & Taxonomy
The skill organizes its intelligence through a standardized file structure and metadata notation:
- Plan Files: Individual markdown files named
[slug].plan.mdcontaining task lists and status headers. - Patterns File: A centralized
patterns.mdfile that stores distilled lessons on sizing and dependency logic. - Task Metadata Table:
| Notation | Meaning | Description |
|---|---|---|
| (S/M/L) | Sizing | Estimates duration from <30m to 2h+ |
| (depends: N) | Dependency | Tasks that must be completed before this one |
| ↳ parallel | Concurrency | Tasks that can run simultaneously with others |
| [x] | Completion | Marked with a brief summary of the result |
Task Decomposition Advanced Features
- Recursive Learning Loop: Automatically updates a patterns library to refine task sizing and avoid repeating dependency mistakes.
- Sub-agent Delegation: Supports spawning sub-agents for independent branches of work while tracking their session IDs in the master plan.
- Resumption Logic: Specifically designed to scan existing plans after a restart to find the first unblocked task and resume work seamlessly.
- Size-Aware Reporting: Generates status updates that reflect the actual weight of remaining work rather than just a simple task count.
- Failure Recovery: Provides a structured way to abandon plans or mark partial progress, ensuring value is captured even if a project is unrecoverable.
SKILL.md
Loading
Related Openclaw Skills
Church Account Automation (LDS/LCR)
An automation tool for managing LDS Church account tasks, Leader and Clerk Resources (LCR), and member directories efficiently.
ClawMart Creator
The ClawMart Creator skill enables developers to manage their marketplace presence and publish Openclaw Skills directly through conversational commands.
ClawMart Browser
A specialized tool for searching and evaluating AI personas and skills directly within the ClawMart marketplace.
Second Brain
A personal knowledge base powered by Ensue for capturing and retrieving structured understanding that compounds over time.
OpenClaw Health Brief
A comprehensive automation skill for aggregating and normalizing health data from Oura, WHOOP, and Withings into a unified daily status report.
Openclaw Health Brief
A comprehensive automation skill that aggregates health metrics from Oura, Whoop, and Withings into a single, actionable daily brief.
Featured*
