Task Director is a sophisticated workflow management skill that treats complex AI agent tasks like movie storyboards, ensuring planning happens before execution.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install task-director
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-director 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 Director?
Task Director is a powerful orchestration framework designed for AI agents within the Openclaw Skills ecosystem. It moves away from the 'run and pray' approach of standard agents by introducing a cinematic metaphor for task management. Users and agents collaborate to create a detailed script composed of scenes and shots, allowing for human oversight and approval before any commands are actually executed.
By structuring workflows into a formal storyboard, Task Director provides a robust safety net for automation. It features native support for fallback actions, allowing an agent to automatically pivot to an alternative command if the primary action fails. This structured approach ensures higher reliability in complex deployments, environment setups, and multi-stage computational tasks.
Task Director Use Cases
- Orchestrating multi-stage full-stack project deployments where environment checks are critical.
- Managing complex CI/CD pipelines that require manual approval gates between stages.
- Handling automated server migrations with built-in fallback recovery steps.
- Executing sequential data processing tasks that need to be paused, skipped, or retried based on real-time output.
How Task Director Works
- Script Creation: The AI agent or user generates a
scenes.jsonfile defining the scenes (stages) and shots (individual actions). - Drafting: The skill initializes the task in a draft state, allowing for review and modification of commands and fallbacks.
- Approval: The user reviews the storyboard and issues an approve command, transitioning the state from draft to approved.
- Action Loop: The agent requests the next shot via the action command, executes the logic, and reports the result back to the director.
- Fallback Handling: If a shot fails (outcome: ng), Task Director automatically identifies if a fallback action exists and provides it for the next execution attempt.
- Wrap Up: Once all scenes are completed, the wrap command calculates final statistics, including total duration and success rates.
Task Director Setup
To use Task Director, ensure you have Python installed and the skill added to your environment. You can interact with it using the director.py script.
# Define the alias for convenience
DIRECTOR="python scripts/director.py"
# Create a new storyboard from a JSON configuration
$DIRECTOR create --title "My Complex Task" --scenes scenes.json
# Approve the script to begin execution
$DIRECTOR approve
Task Director Data Schema & Taxonomy
Task Director organizes data within the Openclaw Skills structure under ~/.openclaw/memory/movies/. Each task is stored as an individual JSON file containing the following taxonomy:
| Entity | Attributes |
|---|---|
| Scene | Name, Description, Estimated Time, List of Shots |
| Shot | Action, Command, Fallback Action/Command, Status (pending, running, done, ng, skipped) |
| Metadata | Movie ID, Title, Current State (draft, approved, filming, wrapped), Execution Logs |
Task Director Advanced Features
- Granular Control: Use the 'cut' command to pause operations and 'action' with specific scene/shot IDs to jump to any part of the script.
- Intelligent Fallbacks: Define alternative commands directly within the shot schema to handle predictable errors without failing the entire workflow.
- Step Skipping: Skip unnecessary steps with a documented reason to maintain a clean audit trail while saving time.
- Comprehensive Analytics: The 'wrap' command provides a post-mortem summary of the entire task lifecycle, perfect for performance optimization in Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
Plan First
A structured problem-solving skill that forces AI agents to design detailed blueprints before executing complex, multi-step tasks.
Pattern Recognition Skill
An intelligent learning system that identifies operational patterns and generates optimized templates to enhance agent performance.
Exhibitor Show History
A specialized research tool that retrieves the complete trade show exhibition history for any company using the Lensmor API.
Autonomous Research Loop
An infinite, self-driven research agent that autonomously identifies topics, conducts deep investigations, and documents findings in Feishu.
Aliyun ESA Manage
A comprehensive automation skill for managing Alibaba Cloud Edge Security Acceleration (ESA) services including static deployments, serverless edge functions, and distributed storage.
Markdown to NanoBanana PPT
An intelligent workflow that converts Markdown documents into visual-rich PowerPoint presentations using AI image generation.
Featured*
