agent-control

What This Skill Does

The agent-control skill acts as a natural language interface for managing OpenClaw isolated agents. Instead of memorizing complex CLI flags and syntax, you can issue conversational commands to create, list, modify, or remove agents directly from your chat interface. It acts as a bridge between high-level user intent and the low-level openclaw binary, handling path defaults, model selection, and channel routing automatically.

Installation

To integrate this skill into your environment, run the following command in your terminal: clawhub install openclaw/skills/skills/alincatalin/agent-control

Use Cases

This skill is designed for power users who maintain multiple agent personas or distinct workspaces. Use it when you need to:

• Rapidly prototype new agents by defining custom models and workspace directories.
• Manage multi-channel connectivity by binding or unbinding specific chat channels to different agent instances.
• Perform administrative cleanup by listing and deleting unused agents to save system resources.
• Personalize your agent interfaces by updating identities, emojis, and avatars on the fly.

Example Prompts

1. "Create a new agent named 'ResearchBot' using the 'gpt-4o' model in my projects folder."
2. "Switch my current channel to the 'SupportAgent' and set its display name to 'HelpDesk'."
3. "List all my current agents, then delete the one named 'TestSandbox'."

Tips & Limitations

• Safety First: The agent delete command is intentionally destructive. The skill requires an explicit confirmation step before it proceeds to execute the CLI deletion, preventing accidental loss of agent data.
• Ambiguity: If you issue an incomplete command (e.g., forgetting the channel ID), the agent will prompt you with a single, focused question to resolve the ambiguity rather than guessing or failing silently.
• Scope: This skill is scoped specifically to the openclaw CLI ecosystem. It does not provide general system shell access. If a command fails, the skill provides the raw error message along with a concrete, actionable fix to help you get back on track. For advanced automation, the skill leverages scripts/example.py as a deterministic parsing layer to ensure command execution is consistent.