Browse reusable automation Skills, installation commands, and recommended use cases.

We maintain two Skills for image generation and public Feishu table monitoring. Start with the web app for everyday use; use the commands below for agents and automation.

2026-11-03

1. Currently Supported Skills

Skill	Version	Purpose
`guishu-gpt-image-2`	1.1.0	Text-to-image, reference-image editing, masked edits, and recovery of existing image downloads.
`feishu-public-table-monitor`	1.0.5	Detects changes in public Feishu prices, multipliers, and model lists, and drafts notifications.

Installation requires Node.js 22+. The image script requires Python 3.10+ and no pip dependencies. The default workspace is ~/.openclaw/workspace; replace the path for a custom workspace.

2. guishu-gpt-image-2: gpt-image-2 Image Generation Skill

Verify that the ClawHub publisher is quqi1599. The current CLI takes the Skill name without a publisher prefix. For an existing installation, use update below; back up local edits instead of adding --force by default.

npx --yes clawhub@latest --workdir ~/.openclaw/workspace install guishu-gpt-image-2 --version 1.1.0
npx --yes clawhub@latest --workdir ~/.openclaw/workspace update guishu-gpt-image-2 --version 1.1.0

Configure the Key and check locally

The script reads LLM_TOKEN_API_KEY, then GUI_SHU_TOKEN_KEY. Version 1.1.0 no longer automatically uses OPENAI_API_KEY; if that variable holds your LLM API Gateway Key, explicitly add --key-env OPENAI_API_KEY. Do not paste keys into chat.

export LLM_TOKEN_API_KEY="YOUR_GUI_SHU_TOKEN"
python3 "$HOME/.openclaw/workspace/skills/guishu-gpt-image-2/scripts/generate.py" \
  --prompt "Blue smart speaker product poster, clean background, no invented logos or prices" \
  --size 1024x1024 --quality high --n 1 --dry-run

--dry-run checks local inputs without network access, Key lookup, or file writes. Remove it only when ready to generate a billable image. Empty prompts are rejected.

Windows PowerShell

py -3 must select Python 3.10+; use python if the py launcher is unavailable. These environment variables apply only to the current terminal; a separately launched agent needs its own environment configuration.

npx --yes clawhub@latest --workdir "$env:USERPROFILE\.openclaw\workspace" install guishu-gpt-image-2 --version 1.1.0
$env:LLM_TOKEN_API_KEY = "YOUR_GUI_SHU_TOKEN"
$imageSkill = Join-Path $env:USERPROFILE '.openclaw\workspace\skills\guishu-gpt-image-2\scripts\generate.py'
py -3 $imageSkill --prompt "Blue smart speaker product poster, clean background, no invented logos or prices" --dry-run

Reference-image editing and download recovery

Repeat --image for multiple local reference images. Add --mask ./mask.png for a masked edit: the first image and mask must be same-size PNGs, and the mask needs an Alpha channel. Transparent areas mark the edit region. Each image is limited to 50 MiB.

python3 "$HOME/.openclaw/workspace/skills/guishu-gpt-image-2/scripts/generate.py" \
  --image ./product.png \
  --prompt "Change only the background to a warm gray studio; preserve product shape, colors, and logo" \
  --out-dir ./outputs/product-edit

If generation succeeded but downloading failed, --resume retrieves only existing results, without a Key or a new generation request. It does not submit unstarted batch jobs. Recovery may be impossible without a saved response or after the original URL expires. Check usage after a timeout; do not rerun the entire batch.

python3 "$HOME/.openclaw/workspace/skills/guishu-gpt-image-2/scripts/generate.py" --resume ./outputs/product-edit --json

Output files and model selection

Images use numbered filenames. state.json tracks each job, response-NNN.json enables recovery, request.json summarizes parameters, and index.html previews local results. Prompts are not saved unless you add --save-prompts. Start new jobs in an empty directory; one failed image does not block other valid images in the same response.

The default remains gpt-image-2; explicitly select gpt-image-2.5-flare or gpt-image-2.5-sunburst when needed. Access and charges depend on your account. Generation/download socket timeouts default to 1200/120 seconds and are configurable, not completion-time guarantees.

ClawHub · Full image setup guide · Web app

3. feishu-public-table-monitor: Public Feishu Table Monitoring Skill

feishu-public-table-monitor monitors publicly accessible Feishu Wiki and document pages. It locates a pricing or model table near a specified section, detects version date changes, added or removed models, multiplier changes, and pricing changes, then formats them as Markdown suitable for customer notifications or Telegram messages.

What You Can Use It For

Monitor model pricing tables for multipliers, prices per 1 million Tokens, and model availability.
Monitor model lists for additions, removals, and naming convention changes.
Monitor public product tables for changes to plans, pricing, product lists, and promotions.
Add it to a scheduled task that stays silent when nothing changes and generates a notification when a change is detected.

Installation

npx --yes clawhub@latest --workdir ~/.openclaw/workspace install feishu-public-table-monitor --version 1.0.5

Quick Start Example

For the first run, print a snapshot to confirm that the correct table was captured.

Before running the command, enable public internet sharing for the target document in Feishu. Then replace PASTE_YOUR_PUBLIC_FEISHU_DOC_URL_HERE below with its public Feishu URL. Links that require sign-in cannot be monitored.

FEISHU_PUBLIC_DOC_URL='PASTE_YOUR_PUBLIC_FEISHU_DOC_URL_HERE'
python3 "$HOME/.openclaw/workspace/skills/feishu-public-table-monitor/scripts/monitor_feishu_price_table.py" \
  "$FEISHU_PUBLIC_DOC_URL" \
  --section-title 'Model pricing and billing' \
  --print-snapshot

After confirming the result, run it again to establish a baseline. On future scheduled runs, it outputs NO_REPLY if nothing has changed. If it detects a change, it outputs a formatted Markdown notification.

FEISHU_PUBLIC_DOC_URL='PASTE_YOUR_PUBLIC_FEISHU_DOC_URL_HERE'
python3 "$HOME/.openclaw/workspace/skills/feishu-public-table-monitor/scripts/monitor_feishu_price_table.py" \
  "$FEISHU_PUBLIC_DOC_URL" \
  --section-title 'Model pricing and billing' \
  --title 'LLM API Gateway model pricing update'

Recommended Workflow

Use --print-snapshot to confirm that the correct table was captured.
Run the Skill normally once to create a local baseline snapshot.
Add it to cron, an OpenClaw scheduled task, or your own Agent workflow.
Do not send anything when it outputs NO_REPLY. When it outputs Markdown, send that message directly to customers or group chats.

Model list: Model list
Model pricing and billing: Pricing guide

What this automates: Whenever model prices, multipliers, or model availability changes, your Agent can detect the update, format it, and send a reminder automatically, reducing the risk of missing important updates during manual checks.

4. Using the Skills After Installation

Goal	Recommended Skill	How to Use It
Generate images	`guishu-gpt-image-2`	Describe the image in a prompt and let the Agent call the script to generate it.
Monitor pricing tables	`feishu-public-table-monitor`	Provide a public Feishu URL and section title, then schedule checks for changes.
Notify customers	`feishu-public-table-monitor`	When a change is detected, pass the Markdown output to a Telegram, Feishu, or WeChat sending workflow.

Return to the main guide for more information about models, pricing, and API access.

Your privacy choices

LLM API Gateway Skills