NovelAI OpenClaw Adaptor

Use this skill to explain how a local adaptor can bridge OpenClaw's OpenAI-style API calls to NovelAI, and how to configure OpenClaw to talk to that local endpoint.

Scope

This skill is for explanation and local configuration guidance:

• - Explain that the adaptor is a local proxy/shim, not a hosted service.
• Explain how base_url and model names should be configured in OpenClaw.
• Help the user choose a supported text model or image model.
• Describe the local commands a user may run after they have verified the package source.

Do not use this skill to:

• - Auto-install software.
• ask the user to paste secrets into chat.
• present unverified third-party packages as implicitly trusted.

Safety rules

Follow these rules whenever this skill is used:

1. 1. Treat installation as optional and approval-based.
2. Before suggesting installation, tell the user to verify the package source, maintainer, and repository or project page.
3. Prefer local source checkout or an already-verified package source when available.
4. Never ask the user to paste a NovelAI API key into chat.
5. Never include secrets inline in command examples.
6. If the package source cannot be verified, stop at configuration guidance and ask the user how they want to proceed.

Verification-first workflow

If the user wants to enable runtime use, follow this order:

1. 1. Check whether the adaptor is already present locally or already installed.
2. If it is not present, explain that the package source should be verified before any installation.
3. Ask for approval before running any install command.
4. After the user has verified the source and approved installation, use the package's documented install method.
5. Prefer interactive or local-only credential entry during configuration.

Safe examples:

CODEBLOCK0

CODEBLOCK1

Do not use examples like:

CODEBLOCK2

If the user needs help deciding whether the package is trustworthy, suggest reviewing its repository, release history, and maintainers before installation.

Supported models

Text models:

• - INLINECODE1
• INLINECODE2
• INLINECODE3
• INLINECODE4
• INLINECODE5
• INLINECODE6
• INLINECODE7
• INLINECODE8
• INLINECODE9

Image models:

• - INLINECODE10
• INLINECODE11
• INLINECODE12
• INLINECODE13
• INLINECODE14
• INLINECODE15

When helping with setup, ask the user which model they want instead of assuming silently. If they do not care, recommend:

• - Text: INLINECODE16
• Image: INLINECODE17

OpenClaw configuration

When explaining how to connect OpenClaw to the adaptor:

1. 1. Set base_url to the local adaptor endpoint, such as http://127.0.0.1:xxxx/v1 or http://localhost:xxxx/v1.
2. Set the OpenClaw model name to the adaptor-exposed model the user selected.
3. Clarify that credential handling belongs to the local adaptor configuration, not the chat.
4. If OpenClaw insists on an API key field, explain that some clients accept a placeholder value such as sk-local, but the real NovelAI credential should stay in the local adaptor config only.

Helpful local commands

If the user has already verified the package source and approved local usage, these commands are relevant:

CODEBLOCK3

Use novelai-config init as the normal guided setup entry point. It should collect local configuration such as:

• - UI language
• NovelAI credential through local input
• Default shim model
• Default image output directory
• Default image model

Image generation usage

Once the local adaptor is configured and running, image generation can use the normal OpenClaw or OpenAI-style prompt flow.

Example prompt:

INLINECODE23

The adaptor is responsible for translating that prompt into the format expected by NovelAI.