How to Install and Use md2wechat-skill in OpenClaw

OpenClaw is where people most often mix up two different installation stories.

They install the skill shell and assume the runtime now exists.

That assumption is exactly what breaks the workflow.

The mental model that keeps OpenClaw sane

For OpenClaw, think about two separate pieces:

• the OpenClaw skill shell
• the md2wechat CLI runtime

You need both.

If one is present and the other is missing, the integration is still broken.

That is why OpenClaw should never be documented as if it were the same thing as the Claude Code or Codex path.

The fastest complete installation path

The most direct way to install the OpenClaw version is:

curl -fsSL https://github.com/geekjourneyx/md2wechat-skill/releases/download/v2.0.7/install-openclaw.sh | bash
export PATH="$HOME/.local/bin:$PATH"

That path exists for a reason: it is the easiest way to end up with both the skill package and the CLI runtime in a consistent version.

If you only want the CLI in your environment first and already have Node and npm, you can also run:

npm install -g @geekjourneyx/md2wechat

That installs only the CLI. It does not create the OpenClaw skill shell under ~/.openclaw/skills/md2wechat/.

The ClawHub path

If you already live in ClawHub, you can install the skill shell with:

npx clawhub@latest install md2wechat

That can be a good shell-level install path, but it does not guarantee that the md2wechat CLI is also available.

So after ClawHub, you still need to check:

command -v md2wechat

If there is no output, the runtime is still missing.

What to verify right after install

These checks should happen before any publishing task:

md2wechat version --json
md2wechat config init
md2wechat config validate
md2wechat capabilities --json
ls ~/.openclaw/skills/md2wechat/

That tells you:

• the CLI is callable
• config files can be initialized
• discovery works
• the skill shell exists where OpenClaw expects it

The first task worth running

Once verification is clean, start with a plain preview task:

md2wechat convert article.md --preview

Then move up to the more sensitive parts:

md2wechat convert article.md --draft --cover cover.jpg

That second step is where missing WeChat credentials or incomplete config usually surface.

The most common OpenClaw mistakes

1. Treating ClawHub install as the full story

It is not always the full story. It may install the shell without the runtime.

2. Debugging the wrong directory

The OpenClaw path is:

~/.openclaw/skills/md2wechat/

Do not debug ~/.claude/skills/ and assume the result applies to OpenClaw.

3. Forgetting that API mode is still the default

Even inside OpenClaw, convert defaults to API mode unless you explicitly pass --mode ai.

4. Expecting drafts to work without credentials

OpenClaw does not remove the need for:

• WeChat AppID
• WeChat Secret
• usually an md2wechat.cn API key for API-mode workflows

A useful prompt for OpenClaw

If you want the agent to do the setup work, use:

Please install the OpenClaw version of md2wechat, verify the skill shell and CLI runtime, then run md2wechat capabilities --json before doing any conversion work.

That prompt forces the runtime checks to happen before the publishing promises.

Where to continue

• OpenClaw
• md2wechat-skill Guide
• md2wechat-skill FAQ