A Fail-Closed scheduling layer for Openclaw Skills designed to manage agent concurrency, prevent network conflicts, and enforce granular permission auditing.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-concurrency-controller
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 agent-concurrency-controller 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 Agent Concurrency Controller?
The Agent Concurrency Controller is a sophisticated scheduling framework built for Openclaw Skills, inspired by the three-layer architecture of Claude Code (Tool, ToolUse, and Task). It serves as a safety-first coordination layer that prevents common pitfalls in multi-agent environments, such as network session collisions and token explosions caused by unmanaged sub-agent outputs. By default, the controller operates on a Fail-Closed principle, meaning all tasks are queued serially unless explicitly marked as safe for parallel execution.
This skill is essential for developers building complex autonomous systems where reliability and security are paramount. It categorizes operations into tiered sensitivity levels—normal, sensitive, and critical—ensuring that high-risk actions like financial transactions or public deployments require explicit human confirmation while routine research and search tasks proceed with automated efficiency.
Agent Concurrency Controller Use Cases
- Preventing isolated session network conflicts when running multiple cron jobs or sub-agents simultaneously.
- Auditing sensitive file system operations or configuration changes with mandatory logging.
- Managing priority-ordered task execution where specific agents must finish before others start.
- Controlling API costs and context window usage by synthesizing worker results through a parent coordinator.
- Implementing human-in-the-loop checkpoints for critical business logic within Openclaw Skills workflows.
How Agent Concurrency Controller Works
- Incoming tasks are submitted to the Task Queue layer where they are sorted by priority (1-10) and dependency topology.
- The ToolUse Context layer intercepts the task to verify permissions and check the isConcurrencySafe flag.
- If a task is marked as critical, the controller pauses and logs a request for user confirmation before proceeding.
- The agent is spawned in either an isolated or main session depending on the required context continuity.
- Upon completion, the worker agent sends a status signal to the coordinator, which synthesizes the final result for the user.
Agent Concurrency Controller Setup
To deploy the controller, integrate the implementation script into your project and configure the permission manifest within your Openclaw Skills environment.
Verify the deployment by checking the queue status:
python skills/agent-concurrency-controller/agent_concurrency_controller.py
Ensure your agent has the following permissions enabled in your configuration: agent:spawn, file:write, and log:append.
Agent Concurrency Controller Data Schema & Taxonomy
The controller manages its state and audit trails using a structured JSON and log-based schema:
| File Path | Description | Format |
|---|---|---|
memory/agent-queue-state.json |
Persistent storage for pending and running tasks. | JSON Object |
logs/agent-concurrency.log |
Records task lifecycle events and queue depth. | JSON Lines |
logs/sensitive-operations.log |
Audit trail for sensitive and critical level actions. | JSON Lines |
agent_concurrency_controller.py |
Core logic and Python API interface. | Python Script |
Agent Concurrency Controller Advanced Features
- Priority-based task preemption allowing high-importance agents to skip the queue.
- Topology-aware dependency management for complex, multi-stage agent workflows.
- Circuit-breaking mechanisms that timeout stalled agents to prevent queue blockages.
- Multi-level operation auditing (normal, sensitive, critical) for enterprise-grade security.
- Seamless synthesizer pattern that aggregates sub-agent data to prevent token-heavy direct output.
SKILL.md
Loading
Related Openclaw Skills
Session Context Injector
A specialized skill that ensures AI agents maintain project continuity in Telegram by injecting context from local status files after session resets.
Tool Result Size Controller
An optimization utility for AI agents that prevents context window overflow by automatically offloading large tool outputs to disk storage.
Neural Learning Engine
A lightweight AI module that simulates a neural network learning loop to enable pattern recognition and adaptive responses in intelligent agents.
Safe File Editor
A robust file editing utility for AI agents that implements atomic operations, mandatory diff previews, and automated rollback recovery.
Context Cache Manager
A sophisticated context management tool for Openclaw Skills that ensures fork-safe state inheritance and prevents token explosion through intelligent compression.
Chill Beijing
A real-time Beijing activity recommender that scrapes top entertainment platforms to suggest after-work and weekend relaxation options.
Featured*
