An agent-ready deployment skill for turning local static HTML into a stable public Cloudflare Pages URL.
HTML Deploy is built for agents that should work from local files, CLI, and verification instead of clicking through a web app.
Compatible targets:
- Codex
- Claude Code
- Hermes
- OpenClaw
- Install the skill into your local agent skills directory.
~/.codex/skills/html-deploy- Trigger it with a direct prompt.
Use $html-deploy to publish the HTML in ./dist and give me the stable production URL.
- Expect the agent to deploy with Cloudflare Pages, verify the live page, and return a stable URL like:
https://your-project.pages.dev
HTML Deploy gives an agent one clean job:
- take a local static site
- publish it to Cloudflare Pages
- keep the public URL stable when needed
- verify the live result before reporting success
- Publishes a local HTML/CSS/JS folder to Cloudflare Pages
- Reuses an existing Pages project when the public URL should stay unchanged
- Explains and automates an OpenDesign-style no-UI deployment flow
- Guides users through Cloudflare API Token setup without exposing secrets
This repository is packaged as a SKILL.md-based skill, so the most direct drop-in experience is for agent systems that support skill folders and markdown instructions.
In practice:
- Codex can use the repository directly as a local skill
- Claude Code, Hermes, and OpenClaw can reuse the same instructions, references, prompts, and workflow design
- if a target framework uses a different plugin or skill manifest format, keep
SKILL.mdas the source of truth and adapt the wrapper only
Use this skill when you want an agent to handle requests like:
- "把这个 HTML 部署成一个可分享链接"
- "帮我把这个静态页面上线到 Cloudflare Pages"
- "不要打开 UI，直接把页面发出去"
- "更新这个
pages.dev项目，但不要换链接"
Copy or symlink this repository into your Codex skills directory.
Common location:
~/.codex/skills/html-deployFor Claude Code, Hermes, and OpenClaw:
- place this repository in the runtime's skill, prompt, or agent-capability directory
- use
SKILL.mdas the primary instruction file - keep the
references/folder next to it so the agent can load deeper guidance on demand - adapt only the runtime-specific wrapper if needed; do not fork the workflow content unless behavior truly differs
If you want another agent to install this skill for you, send it this prompt:
Install the skill from https://github.com/kelin3296-jpg/html-deploy into my local agent skills directory. Clone the repository, keep SKILL.md and the references folder together, verify that SKILL.md and agents/openai.yaml exist, and confirm the skill can be discovered or referenced locally by my agent runtime.
If the target agent supports symlinks and you prefer that setup, replace the clone step with a symlink into the same destination.
After installation, ask the agent for the task directly, or invoke the skill by name if the runtime supports explicit skill calls.
Examples:
Use $html-deploy to publish the HTML in ./dist and give me the stable production URL.
Use $html-deploy to update my existing Cloudflare Pages project without changing the public link.
User:
Use $html-deploy to publish the static site in ./site.
Project name should be demo-landing-pages.
If Cloudflare credentials are missing, tell me exactly what I need to apply for.
Agent:
I’ll deploy the local static site in ./site to Cloudflare Pages, reuse the project name demo-landing-pages, and return only the stable production URL after verification. If credentials are missing, I’ll stop and tell you how to create the required Cloudflare API Token and Account ID setup.
User:
Use $html-deploy to redeploy this existing project, keep the same pages.dev URL, and verify the new copy is live.
Agent:
I’ll redeploy into the existing Pages project so the production URL stays unchanged, then check the live page before reporting completion.
This skill assumes:
- you have a local static site directory with
index.html wrangleris available, or can be installed- Cloudflare credentials are available at runtime through environment variables or a protected local config
See references/cloudflare-token-setup.md for a safe token application flow.
Cloudflare API Token page:
https://dash.cloudflare.com/profile/api-tokens
Recommended custom token permission:
Account -> Cloudflare Pages -> Edit
Common environment variables:
export CLOUDFLARE_API_TOKEN="..."
export CLOUDFLARE_ACCOUNT_ID="..."Do not commit real credential values, .env files, or machine-specific config files into the repository.
This skill is opinionated about three things:
- Prefer local files and CLI over clicking through a deployment UI
- Return the stable production URL, not a one-off deployment preview link
- Verify the page after deployment before declaring success
html-deploy/
├── SKILL.md
├── README.md
├── LICENSE
├── .gitignore
├── agents/
│ └── openai.yaml
└── references/
├── cloudflare-token-setup.md
├── examples.md
├── github-publishing.md
└── workflow.md
- SKILL.md: agent-facing instructions
- TESTING.md: static validation and live deployment test flow
- references/workflow.md: deployment flow and validation
- references/cloudflare-token-setup.md: token application and safe local setup
- references/examples.md: request and response examples
This project is not a Cloudflare Pages wrapper for one specific coding agent. It is a reusable deployment workflow package for AI agents that generate or edit HTML locally and then need to publish it with a safe, repeatable, stable-link workflow.
MIT