Quests

What This Skill Does

The Quests skill provides a powerful, structured framework for OpenClaw agents to manage long-term, multi-step processes. Unlike ephemeral chat logs or scattered memory files, Quests creates a centralized, persistent record of complex procedures. It treats a task as a stateful object, tracking progress, metadata, decisions, contacts, and specific substeps. The skill operates via a clean CLI that allows the agent to break down daunting bureaucratic, legal, or technical hurdles into bite-sized, manageable actions, ensuring the human user never feels overwhelmed.

Installation

To integrate this skill into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/poloio/quests

Ensure that your workspace directory has write permissions, as the skill maintains a quests.json file to persist your data across sessions. Once installed, the quest alias becomes available for immediate use.

Use Cases

Quests is designed for processes that span multiple days or sessions. Ideal scenarios include:

• Legal/Bureaucratic Procedures: Tracking visa applications, tax filings, or permit renewals where documentation and dates are critical.
• Technical Projects: Managing multi-step software deployments, home server migrations, or hardware repair projects.
• Complex Planning: Organizing travel itineraries, event planning, or house-hunting efforts that involve contacting multiple stakeholders and verifying complex requirements.
• Knowledge Retention: Capturing "decisions" and "learnings" during a project so that the agent maintains context, preventing the loss of critical project logic.

Example Prompts

• "I need to get my driver's license renewed; it's a long process, can we create a quest for it?"
• "What is the next step in my home renovation quest, and do I have any pending contacts I need to call?"
• "Update my progress: I finally finished the permit application step of the building project."

Tips & Limitations

• Context Management: The quest context command is your best friend. Use it at the start of a session to load the full state into the agent's memory window, avoiding the need to parse historical message logs.
• Don't Over-engineer: If a task can be done in one afternoon, it's probably better handled as a simple todo list. Quests are for processes that live in the long-term.
• Fuzzy Matching: You don't need to type the full quest name; the CLI supports partial matches, making it fast to interact with during a conversation.