deploy-agent

What This Skill Does

The deploy-agent is a comprehensive lifecycle management tool designed for the OpenClaw AI agent ecosystem, streamlining the transition from local development to production. It orchestrates a structured, multi-step pipeline consisting of project initialization, automated builds, local validation testing, source control integration via GitHub, and final deployment to Cloudflare Pages. By enforcing human-in-the-loop approvals at every critical transition point, this skill ensures that infrastructure changes and deployments are intentional and verified. The agent maintains persistent state files, allowing it to resume deployment workflows after interruptions or session timeouts.

Installation

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

clawhub install openclaw/skills/skills/sherajdev/deploy-agent

Ensure that you have the required prerequisites installed on your system: gh (GitHub CLI) for repository operations, wrangler for Cloudflare interaction, git for version control, and jq for state file management. These tools are essential for the agent to execute its workflows successfully.

Use Cases

This skill is ideal for solo developers, prototyping agencies, and teams building full-stack web applications who require a standardized deployment process. It is particularly useful for rapid MVP iteration, where setting up continuous deployment pipelines might be overkill, but manual verification of each stage is necessary to avoid breaking production environments. It also serves as a pedagogical tool for junior developers, enforcing best practices like local testing before deployment.

Example Prompts

1. "Initialize a new deployment for my-project and guide me through the design phase."
2. "Check the status of my-blog deployment; what is the next step I need to approve?"
3. "The tests passed for my-app, please proceed to push the code to a new GitHub repository."

Tips & Limitations

• Safety First: Always perform manual code audits before providing the approval to move from the 'test' phase to the 'push' phase, as this pushes code to a public or private repository.
• State Management: The agent relies on JSON files in ~/.clawdbot/skills/deploy-agent/state/. Do not edit these files manually unless you are troubleshooting a stuck deployment.
• Naming: Deployment names must be unique to avoid collisions in the configuration directory.
• Dependencies: If the deployment fails, verify that your local environment is authenticated with both GitHub (gh auth login) and Cloudflare (wrangler login).