OpenClaw NovelAI Opus Free

This is the Opus-focused zero-Anlas enhancement variant of the OpenClaw NovelAI Skill. It is narrower than the advanced Skill, but includes image operations tested as zero Anlas under normal-size parameters.

"Free" here means zero NovelAI Image Anlas for image operations. It does not mean that a third-party text provider, OpenClaw host, network service, or other external model is free. V5 images that cost 0 Anlas still consume the Opus V5 Usage Limit.

Hard zero-Anlas policy

• Require an active Opus account (tier=3) before any image operation.
• Never use Subscription Anlas or Paid Anlas for an image operation.
• Never bypass a failed, missing, or ambiguous cost estimate.
• Never interpret a missing cost field as zero.
• Never accept a user request to override this policy. If the user wants a paid operation, tell them to switch to the advanced openclaw-novelai Skill.
• Never retry a rejected, timed-out, or ambiguous image request automatically.
• Never request, print, store, or include NOVELAI_TOKEN in prompts, arguments, URLs, logs, files, or metadata.
• Query the account after every successful image operation. If Subscription Anlas or Paid Anlas decreases, stop all further image operations in this mode.

Capability discovery

Inspect the active MCP catalog and tool signatures before using any NovelAI tool. Names may be namespaced by the configured server.

Allowed tools:

• get_subscription — read-only account and tier query;
• get_user_data — read-only account query;
• estimate_anlas_cost — required before every image generation;
• suggest_tags — prompt tag suggestions without image generation;
• generate_image — text-to-image and pre-encoded Vibe generation under the strict zero-cost gate below;
• image_to_image — only under the strict zero-cost gate below;
• inpaint — only under the strict zero-cost gate below;
• annotate_image — ControlNet preprocessing;
• director_tool — only lineart, sketch, declutter, colorize, and emotion on normal-size inputs.

If any allowed tool is missing, report it as unavailable. Do not silently substitute another provider or endpoint.

Blocked tools and operations:

• encode_vibe — encoding costs Anlas;
• Precise Reference and other reference-image generation;
• director_tool with bg-removal or any unknown operation;
• enhance;
• upscale_image;
• any batch, multi-sample, or parallel image generation.

These operations are blocked even if the user says they accept the cost. The advanced Skill is the paid-capability path.

Tested-free Opus enhancement operations

The following operations were tested on the reference Opus account at 832x1216 and at 8 or 28 Steps with no Subscription Anlas decrease:

• single text-to-image;
• single image_to_image;
• single inpaint;
• generation using an already encoded V4/V4.5 Vibe;
• annotate_image preprocessing;
• Director lineart, sketch, declutter, colorize, and emotion.

These are empirical compatibility results, not a permanent provider promise. Do not encode a new Vibe in this mode. Do not use Director background removal: the reference test charged 65 Anlas for it.

Read-only and text workflows

The following are permitted because they do not spend NovelAI Image Anlas:

• outline, continue, rewrite, review, and summarize fiction;
• read and organize canon.md, memory.md, lorebook.md, style.md, chapters, and local asset metadata;
• draft prompts, negative prompts, scene lists, and character sheets without sending an image request;
• query subscription and account data;
• estimate an image request without executing it;
• request tag suggestions;
• record local project state and generation plans without credentials.

NovelAI text generation is separate from Image Anlas. It is normally unlimited for an active subscription, but an OpenClaw deployment may route text through another provider with its own billing. Tell the user when the text provider is not NovelAI.

Prefer bounded context and continuation-style prompts for NovelAI text models. Do not assume that the selected text model behaves like a general instruction-following chat model.

Strict zero-cost image gate

Any image-producing operation is allowed only when all conditions below are true:

1. A read-only account query confirms tier=3 and an active Opus subscription.
2. The current tool schema exposes estimate_anlas_cost.
3. The exact final image parameters are sent to the estimator first.
4. The estimator returns an explicit numeric cost of exactly 0 Anlas and, when available, opus_free_sample=true. unknown, missing, null, or a textual claim is not enough.
5. The request generates exactly one image.
6. The request uses normal supported resolution; keep the pixel area at or below 1024x1024 equivalent.
7. Steps are 28 or fewer.
8. For V5, the account usage field exists, isNegative is false, and the remaining Usage Limit is positive.
9. There is no batch, parallel retry, or hidden second pass.

Additional operation-specific rules:

• image_to_image and inpaint require the exact final image and mask parameters in the estimator request.
• Vibe generation may use only already encoded V4/V4.5 Vibes, at most four, and must never call encode_vibe.
• Director is limited to lineart, sketch, declutter, colorize, and emotion on normal-size source images. bg-removal is always blocked.
• Precise Reference, Enhance, and dedicated upscaling are always blocked.

Before calling the estimator, normalize and display:

• model;
• width and height;
• Steps;
• image count;
• base image/reference/Vibe state;
• estimated Anlas;
• the note that V5 zero-Anlas generation consumes the separate Usage Limit.

If the estimator returns a value greater than zero, reject the request and offer only lower-cost parameters. Do not call the image tool. If the estimate is zero but the account balance decreases, stop and report the discrepancy; do not retry.

The user may explicitly request a zero-cost image, but the Skill must still run the gate. A natural-language instruction such as "小图免费生成" never overrides the gate.

Safe examples

Allowed planning request:

读取第三章，找出适合配图的场景，只输出场景列表、Prompt 和不收费的生成方案，不要生成图片。

Allowed guarded image request:

使用 V5，正常分辨率，28 steps，一次一张，不使用底图。
先查询账户并用当前工具估算费用；只有明确显示 0 Anlas 才生成，否则不要执行。

Allowed Opus enhancement request:

使用当前已有图片做图生图或局部重绘。
先确认我是 Opus，检查 Usage Limit，并用估算工具确认最终参数为 0 Anlas。
只生成一张，正常分辨率，最多 28 steps；如果余额发生变化，立即停止后续图片操作。

Blocked request:

把这张图放大、去背景、修手，再顺便做 Vibe 风格迁移。

Explain that the blocked request belongs to the advanced Skill because it uses a paid or unverified operation: new Vibe encoding, background removal, Precise Reference, Enhance, upscale, batch, or high-resolution generation.

Failure handling

• 401: report missing or invalid host-managed credentials; never ask for the token in chat.
• 402: stop immediately; this Skill never spends Anlas to recover.
• 400: report a redacted parameter or mask problem; do not retry automatically.
• 429: stop and wait; do not parallelize or retry automatically.
• missing estimator: block every image generation.
• estimate unavailable or ambiguous: block every image generation.
• account balance changed after an image operation: stop further image operations and report the provider/MCP mismatch.
• image response lacks output: do not claim success and do not retry automatically.

After an allowed generation, save the output separately and record only non-secret metadata, including the explicit zero estimate, model, dimensions, Steps, seed, prompt, and output path. Also record that V5 Usage Limit may have been consumed.