GPT-Image-2 for OpenClaw
/install gpt-image-2-toolkit
GPT-Image-2 for OpenClaw
Use this skill when the user wants GPT-Image-2 image generation available through OpenClaw's built-in image_generate tool.
What this skill does
- Installs a plugin package into the global OpenClaw extensions directory
- Registers provider id
hnbc - Exposes model
hnbc/gpt-image-2 - Verifies installation from file, local runtime, and tool perspectives
- Diagnoses the common split-brain case where disk/runtime sees
hnbcbut the running gateway tool view does not - Reminds you that a running gateway may need a restart before the tool sees the new provider
When to use
- "给 OpenClaw 加一个 GPT-Image-2 图像能力"
- "让 image_generate 支持 hnbc/gpt-image-2"
- "修复 No image-generation provider registered for hnbc"
- "把这个能力做成别的 agent / 机器可安装的形式"
- "检查 GPT-Image-2 provider 为什么工具侧看不到"
Core workflow
Install
- Run
scripts/install.sh. - Run
scripts/self-check.sh. - Note: bundled
hnbcunder/usr/lib/node_modules/openclaw/dist/extensions/hnbcis also a valid source; missing~/.openclaw/extensions/hnbcalone does not prove the provider is absent.
Verify
Check both sides:
- Local runtime/plugin registry
image_generate(action="list")
Diagnose mismatch
If local runtime sees hnbc but image_generate list does not:
- explain that the currently running gateway likely still holds an old provider registry
- ask before restarting the gateway
- do not restart without explicit user approval
- after restart, verify again and test a real generation
Deep diagnosis
If anything is unclear, run scripts/diagnose.sh.
If you need extra interpretation guidance, read references/troubleshooting.md.
Recommended validation order
scripts/self-check.shimage_generate(action="list")- If missing from tool list, compare with local runtime using
scripts/diagnose.sh - After install or restart, run
scripts/e2e-check.sh - Test a real generation with
model="hnbc/gpt-image-2"
For next-version release guidance, read references/release-1.0.1.md.
Important compatibility notes
hnbc/gpt-image-2supports generation, not edit.- Supported sizes:
1024x1024,1024x1536,1536x1024 - Supported aspect ratios:
1:1,2:3,3:2 - Do not pass
resolution; HNBC currently rejects resolution overrides. - Auth should come from either:
HNBC_API_KEY, ormodels.providers.hnbc.apiKey, or- an auth profile for provider
hnbc
Files bundled with this skill
assets/plugin/— the plugin package filesscripts/install.sh— installs the plugin into the global OpenClaw extensions dirscripts/self-check.sh— verifies required files and manifest/package basicsscripts/diagnose.sh— prints a fuller runtime diagnosisscripts/e2e-check.sh— runs a post-install validation checklistreferences/troubleshooting.md— quick error-to-fix mapreferences/release-1.0.1.md— next release checklist and publish command
Verification checklist
scripts/self-check.shreturnsSTATUS=okimage_generate(action="list")showshnbc (default gpt-image-2)- generating with
model="hnbc/gpt-image-2"works - no
resolutionparameter is sent during generation
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install gpt-image-2-toolkit - After installation, invoke the skill by name or use
/gpt-image-2-toolkit - Provide required inputs per the skill's parameter spec and get structured output
What is GPT-Image-2 for OpenClaw?
Install, verify, repair, and diagnose GPT-Image-2 support for OpenClaw so agents can use model refs like hnbc/gpt-image-2 through the built-in image_generate... It is an AI Agent Skill for Claude Code / OpenClaw, with 36 downloads so far.
How do I install GPT-Image-2 for OpenClaw?
Run "/install gpt-image-2-toolkit" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is GPT-Image-2 for OpenClaw free?
Yes, GPT-Image-2 for OpenClaw is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does GPT-Image-2 for OpenClaw support?
GPT-Image-2 for OpenClaw is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created GPT-Image-2 for OpenClaw?
It is built and maintained by kevins9527 (@kevins9527); the current version is v1.0.1.