/install happy-img2-direct
Happy IMG2 Direct Skill
Generate real images through an OpenAI-compatible /images/generations endpoint configured in OpenClaw.
Default behavior:
- provider:
happyunlessOPENCLAW_IMAGE_PROVIDERis set - model:
gpt-image-2unlessOPENCLAW_IMAGE_MODELis set - size:
1024x1024 - timeout:
600000msper image - output:
~/.openclaw/generated-images/ - no local fake-image fallback
- no built-in message delivery; send or attach files using your normal OpenClaw/channel tools
Single image
python3 skills/happy-img2-direct/scripts/run.py \
--prompt "A realistic photo of an orange cat sitting by a window, no text, no watermark" \
--task-name "cat-test" \
--no-send
Useful flags:
--promptrequired--task-nameoutput filename prefix--providerprovider key in OpenClaw config, defaulthappy--modelimage model, defaultgpt-image-2--sizedefault1024x1024--timeout-msdefault600000--output-dirdefault~/.openclaw/generated-images--max-attemptsdefault3, maximum5--retry-base-delay,--retry-max-delay,--retry-jitter--rawmarker for callers that intentionally keep the user prompt unchanged--no-sendaccepted for compatibility; this public skill always leaves delivery to the caller
Successful output is JSON containing ok:true, image_path or output, bytes, model/provider, attempt count, and run directory.
Batch images
python3 skills/happy-img2-direct/scripts/batch_run.py @batch.json
Example:
{
"batch_name": "article-covers",
"max_workers": 4,
"timeout_ms": 600000,
"send_to_feishu": false,
"tasks": [
{"task_name": "cover-1", "prompt": "Realistic shop counter photo, no readable text"},
{"task_name": "cover-2", "prompt": "Realistic office desk photo, no readable text"}
]
}
Batch rules:
- bounded concurrency, current hard maximum
4 - each item has its own task directory and logs
- one failed image does not prevent other images from finishing
- final
batch_result.jsonrecords success/failure per task - delivery is disabled in the public version; use OpenClaw/channel tools to send files
Retry behavior
Retries are limited and only used for retryable failures:
- timeout
- upstream failures
- rate limits
- HTTP
408/429/500/502/503/504 - wrapper parse errors
Non-retryable errors, such as invalid requests or auth failures, fail fast with redacted diagnostics.
Safety and publishing notes
This skill intentionally contains no private OpenClaw IDs, no hard-coded user paths, no API keys, and no channel recipient IDs. It reads provider configuration from the local OpenClaw config at runtime.
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install happy-img2-direct - After installation, invoke the skill by name or use
/happy-img2-direct - Provide required inputs per the skill's parameter spec and get structured output
What is Happy IMG2 Direct?
Generate images with an OpenAI-compatible image provider such as happy/gpt-image-2, with retries and bounded batch concurrency. It is an AI Agent Skill for Claude Code / OpenClaw, with 88 downloads so far.
How do I install Happy IMG2 Direct?
Run "/install happy-img2-direct" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Happy IMG2 Direct free?
Yes, Happy IMG2 Direct is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Happy IMG2 Direct support?
Happy IMG2 Direct is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Happy IMG2 Direct?
It is built and maintained by JerryXn (@jerryxn); the current version is v1.0.0.