agent-task-manager

What This Skill Does

The Agent Task Manager provides a robust framework for orchestrating complex, multi-step agent workflows. Unlike standard linear scripts, this skill introduces state persistence and dependency management, enabling your agents to track progress across sessions. It utilizes a Directed Acyclic Graph (DAG) approach to define task dependencies, ensuring that outputs from one specialized agent are seamlessly passed as inputs to the next. The skill also integrates proactive rate-limit management through a dedicated shell wrapper, which tracks timestamps and handles automatic retries to ensure compliance with external API policies. By implementing this skill, you move from simple command execution to building resilient, professional-grade multi-agent systems capable of resuming tasks after interruptions.

Installation

To integrate the Agent Task Manager into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/bobdevibecoder/bobagent-agent-task-manager

Ensure that you have permissions set for the scripts/ directory to allow the execution of molt_task.py and cooldown.sh.

Use Cases

This skill is ideal for complex, time-sensitive, or multi-stage operations. Use it for:

• Financial Auditing: Chaining together contract analysis, financial risk assessment, and reporting.
• Automated Research: Coordinating web scraping, data extraction, and synthesis into a final brief.
• Content Pipeline: Automating a workflow that drafts, audits, and schedules social media posts while respecting platform rate limits.
• Long-running Batch Jobs: Processes that require state persistence to resume after network drops or hardware sleep cycles.

Example Prompts

1. "Initialize a new task workflow named 'MarketPulse' that first runs the FinancialAnalyst agent to pull data, then uses the ContractAuditor to verify security, and finally publishes the findings to Moltbook using the cooldown manager."
2. "Resume the pending state of my 'ComplianceCheck' task. If the previous step failed due to rate limiting, please trigger the retry logic defined in cooldown.sh."
3. "Create a multi-agent system where Agent A performs a web search, Agent B summarizes the findings, and the task manager ensures that the process pauses for 60 seconds between each search request."

Tips & Limitations

• State Management: Always ensure your task_state.json is backed up if managing high-value, long-term workflows. The file contains the current persistence layer for the agent.
• Rate Limiting: While cooldown.sh handles common API limits, always check the specific documentation for the external services you are interacting with (like Helius or Moltbook) to tune your wait intervals appropriately.
• Modular Design: Keep individual task units small and focused. Smaller, single-purpose roles are much easier to debug within the Task Manager's DAG structure than monolithic agent scripts.