@remixmate/cli 0.9.14 → 0.9.16

package/README.md

@@ -296,7 +296,7 @@ remixmate template-registry --list-templates
 ### Quick start
 
 ```
-@skills/gen-image/SKILL.md Generate an image of a panda, 9:16, using gemini 3.1, Chinese-painting style + follow this doc strictly
+@skills/gen-image/SKILL.md Generate an image of a panda, 9:16, using gemini, Chinese-painting style + follow this doc strictly
 @skills/gen-video/SKILL.md Generate a video of a panda running in a bamboo forest, 9:16, 6 seconds, using veo + follow this doc strictly
 @skills/gen-voice/SKILL.md Generate a voiceover introducing panda habits, around 100 words + follow this doc strictly
 @skills/gen-digital-human/SKILL.md Get the digital human list + follow this doc strictly

package/README.zh-CN.md

@@ -232,7 +232,7 @@ remixmate template-registry --list-templates
 ### 快速体验
 
 ```
-@skills/gen-image/SKILL.md 生成一张熊猫的图片，9:16，调用gemini 3.1，国画风 + 严格按该文档执行
+@skills/gen-image/SKILL.md 生成一张熊猫的图片，9:16，调用gemini，国画风 + 严格按该文档执行
 @skills/gen-video/SKILL.md 生成一段熊猫在竹林奔跑的视频，9:16，长度6秒，调用veo + 严格按该文档执行
 @skills/gen-voice/SKILL.md 生成一段语音，介绍熊猫的习性，大概100字左右 + 严格按该文档执行
 @skills/gen-digital-human/SKILL.md 获取数字人列表 + 严格按该文档执行

package/dist/billing.d.ts

@@ -0,0 +1,44 @@
+/**
+ * Billing capture — makes credit consumption visible to whoever ran the skill.
+ *
+ * Credits are deducted server-side and used to be invisible here: a run printed
+ * its image/video URL and nothing else, so the first time a user noticed the
+ * credit system at all was the `insufficient_credits` error after the balance
+ * had already run out. ab-api now attaches a `billing` object to the envelope of
+ * every response that charged something (see core.Success / credits_billing.go);
+ * this module accumulates those across all the calls a single skill run makes —
+ * a gen-image run polls, a gen-video run polls, and each may charge once — and
+ * renders one footer at the end.
+ *
+ * Two outputs, two audiences, both on stdout:
+ *   - a human-readable footer, which is what the agent relays to the user.
+ *   - one `__progress__` line (phase `billing`) — the existing machine protocol,
+ *     already parsed by ab-agent's executor and skipped by the line-scanning
+ *     parsers in render_video.py, so a nested pipeline caller can aggregate a
+ *     whole run's spend without a new stdout contract.
+ *
+ * Both go to stdout rather than stderr (where the auth footer lives) because
+ * ab-agent hands the LLM `result.stdout || result.stderr` (mcp-tools.ts) — a
+ * stderr-only footer would be invisible in the hosted agent, visible only when
+ * a local host like codex shows both streams. Adding stdout lines is safe:
+ * emitProgress already writes NDJSON there, so no consumer can be treating
+ * stdout as a single JSON document.
+ */
+export interface BillingItem {
+    credits: number;
+    bizType?: string;
+    detail?: string;
+}
+/** The `billing` object ab-api attaches to a charged response. */
+export interface Billing {
+    credits: number;
+    balance: number;
+    items?: BillingItem[];
+}
+/** Accumulate one response's billing object; no-ops on responses that charged nothing. */
+export declare function recordBilling(billing: Billing | undefined | null): void;
+/** Everything charged so far in this process, or null when nothing was. */
+export declare function billingSummary(): Billing | null;
+export declare function emitBillingFooter(): void;
+/** Test seam — resets the accumulator between runs in-process. */
+export declare function resetBilling(): void;

package/dist/billing.js

@@ -0,0 +1,76 @@
+/**
+ * Billing capture — makes credit consumption visible to whoever ran the skill.
+ *
+ * Credits are deducted server-side and used to be invisible here: a run printed
+ * its image/video URL and nothing else, so the first time a user noticed the
+ * credit system at all was the `insufficient_credits` error after the balance
+ * had already run out. ab-api now attaches a `billing` object to the envelope of
+ * every response that charged something (see core.Success / credits_billing.go);
+ * this module accumulates those across all the calls a single skill run makes —
+ * a gen-image run polls, a gen-video run polls, and each may charge once — and
+ * renders one footer at the end.
+ *
+ * Two outputs, two audiences, both on stdout:
+ *   - a human-readable footer, which is what the agent relays to the user.
+ *   - one `__progress__` line (phase `billing`) — the existing machine protocol,
+ *     already parsed by ab-agent's executor and skipped by the line-scanning
+ *     parsers in render_video.py, so a nested pipeline caller can aggregate a
+ *     whole run's spend without a new stdout contract.
+ *
+ * Both go to stdout rather than stderr (where the auth footer lives) because
+ * ab-agent hands the LLM `result.stdout || result.stderr` (mcp-tools.ts) — a
+ * stderr-only footer would be invisible in the hosted agent, visible only when
+ * a local host like codex shows both streams. Adding stdout lines is safe:
+ * emitProgress already writes NDJSON there, so no consumer can be treating
+ * stdout as a single JSON document.
+ */
+import { emitProgress } from './progress.js';
+let creditsCharged = 0;
+let latestBalance = null;
+const chargedItems = [];
+/** Accumulate one response's billing object; no-ops on responses that charged nothing. */
+export function recordBilling(billing) {
+    if (!billing || typeof billing.credits !== 'number' || billing.credits <= 0)
+        return;
+    creditsCharged += billing.credits;
+    if (typeof billing.balance === 'number')
+        latestBalance = billing.balance;
+    for (const item of billing.items ?? []) {
+        if (item && typeof item.credits === 'number' && item.credits > 0)
+            chargedItems.push(item);
+    }
+}
+/** Everything charged so far in this process, or null when nothing was. */
+export function billingSummary() {
+    if (creditsCharged <= 0)
+        return null;
+    return { credits: creditsCharged, balance: latestBalance ?? 0, items: [...chargedItems] };
+}
+/**
+ * Emit the footer for what this process charged. Safe to call on the error path
+ * too: a run can fail after a successful (already billed) generation step, and
+ * that spend still needs to be reported.
+ *
+ * Idempotent — only the first call prints, so a handler and the dispatcher can
+ * both reach for it without doubling the number the user sees.
+ */
+let emitted = false;
+export function emitBillingFooter() {
+    if (emitted)
+        return;
+    const summary = billingSummary();
+    if (!summary)
+        return;
+    emitted = true;
+    emitProgress({ phase: 'billing', credits: summary.credits, balance: summary.balance, items: summary.items });
+    const balance = summary.balance.toLocaleString('en-US');
+    const credits = summary.credits.toLocaleString('en-US');
+    process.stdout.write(`\n💳 Charged ${credits} credits · balance ${balance}\n`);
+}
+/** Test seam — resets the accumulator between runs in-process. */
+export function resetBilling() {
+    creditsCharged = 0;
+    latestBalance = null;
+    chargedItems.length = 0;
+    emitted = false;
+}

package/dist/capabilities.d.ts

@@ -19,8 +19,12 @@ export interface ModelConstraints {
     durationDefault?: number;
     refImageMax?: number;
     sizes?: string[];
-    /** aspect-ratio → [width, height] pixel preset (Seedream). */
+    /** aspect-ratio → [width, height] pixel preset (Seedream). Per model: the same
+     *  ratio maps to different pixels on different Seedream variants. */
     sizePresets?: Record<string, [number, number]>;
+    /** total-pixel bounds (width × height); the backend rescales sizes outside them. */
+    pixelMin?: number;
+    pixelMax?: number;
 }
 export interface ModelDescriptor {
     id: string;

package/dist/http.d.ts

@@ -18,6 +18,7 @@
  *   3. process.env.MM_BACKEND_API_URL   — ab-agent's convention (back-compat)
  *   4. https://api.remixmate.com/api    — production default (zero-config)
  */
+import { type Billing } from './billing.js';
 export { SkillError, EXIT } from './errors.js';
 /**
  * Resolve the ab-api base URL the same way for authenticated and device-flow calls.
@@ -49,6 +50,8 @@ export interface MmResponse<T = unknown> {
     code: number;
     msg?: string;
     data?: T;
+    /** Present only on responses that charged credits — see billing.ts. */
+    billing?: Billing;
 }
 /**
  * POST JSON to ab-api and return the parsed business payload.

package/dist/http.js

@@ -18,6 +18,7 @@
  *   3. process.env.MM_BACKEND_API_URL   — ab-agent's convention (back-compat)
  *   4. https://api.remixmate.com/api    — production default (zero-config)
  */
+import { recordBilling } from './billing.js';
 import { resolvePrivToken, NOT_AUTHENTICATED_HINT } from './auth/resolve.js';
 import { attemptAutoLogin } from './auth/auto-login.js';
 import { EXIT, SkillError } from './errors.js';
@@ -120,6 +121,9 @@ export async function mmPost(ctx, pathOrUrl, body, opts = {}) {
     catch {
         throw new SkillError(`❌ failed to parse response, body is not JSON: ${text.slice(0, 200)}`);
     }
+    // Before the code check: a charged response is always code=0 today, but a
+    // partial-failure envelope that still billed must not lose its billing line.
+    recordBilling(parsed.billing);
     if (parsed.code !== 0) {
         // ab-api reports auth failures in the envelope (HTTP 200 + code=401), so the
         // business-code path needs the same 401 → "re-authorize" mapping as above.

package/dist/manifest.json

@@ -1,7 +1,7 @@
 {
   "schemaVersion": 1,
-  "version": "0.9.14",
-  "generatedAt": "2026-08-12T07:37:21.011Z",
+  "version": "0.9.16",
+  "generatedAt": "2026-08-18T05:34:36.073Z",
   "skills": [
     {
       "id": "export-jianying",
@@ -54,7 +54,7 @@
       "tier": "atomic",
       "category": "asset",
       "title": "AI Image Generation",
-      "summary": "AI image generation: produce an image from a text prompt. Supports Seedream and Gemini models, plus image-to-image with reference images.",
+      "summary": "AI image generation: produce an image from a text prompt. Supports the Seedream family (including a high-fidelity 'pro' variant) and Gemini, plus image-to-image with reference images.",
       "triggers": [
         "AI image generation, text-to-image, \"draw me ...\", \"generate an image of ...\"",
         "Image-to-image, reference image, style transfer, image variation",
@@ -247,11 +247,11 @@
       "title": "Web Page Recording",
       "summary": "Drive a headless browser (Playwright Python) to RECORD any URL to a video, then (by default) transcode to mp4, grab a cover frame, upload to VOD and return a playable CDN URL. Modes: fixed-duration recording, condition-triggered stop (element appears / disappears), auto-scroll from top to bottom, custom storyboards, and parameterized templates. Storyboard scenes: highlight / focus / zoom / scroll / virtual-cursor click / type / hover / caption / title-card / arrow / numbered sequence / redact / code-line highlight. For still images (png/jpg) use the web_screenshot tool instead. Use this tool whenever the user wants a video / recording / screencast of a web page: record a page, scroll-through video, page-from-top-to-bottom clip, demo of clicks/typing/hover, storyboard / multi-scene intro video, or a template-based clip.",
       "triggers": [
-        "录屏、网页录制、录制视频、录一段操作、生成 webm、screencast",
-        "滚动录屏、页面从头划到尾的视频",
-        "多场景视频、分镜视频、storyboard、按时间线编排",
-        "演示点击 / 输入 / 悬停（虚拟鼠标 + 涟漪 / 打字机 / 触发 tooltip）",
-        "模板视频一行出片、按顺序编号高亮多个区域并录制"
+        "Screen recording, record a web page, record a video, capture an interaction, produce a webm, screencast",
+        "Scroll recording, a video that pans the page from top to bottom",
+        "Multi-scene video, storyboard video, timeline-sequenced clips",
+        "Animated demos of clicks / typing / hover (virtual cursor + ripple / typewriter / triggering tooltips)",
+        "One-command template clips, numbering and highlighting several regions in sequence and recording it"
       ],
       "entry": {
         "type": "python",
@@ -275,11 +275,11 @@
       "title": "Web Page Screenshot",
       "summary": "Drive a headless browser (Playwright Python) to capture any URL to a local STILL IMAGE (png/jpg): full-page / viewport / element / region screenshots, with device emulation, waiting, hide/mask/redact, and static annotations (highlight / arrow / caption / numbered sequence / redact). Images only — for video / recording / scroll-through / storyboard clips (webm) use the web_record tool instead.",
       "triggers": [
-        "网页截图、网页截屏、整页截图、长截图、full page screenshot",
-        "截某个元素 / 区域、局部截屏、focus 某个区域",
-        "截图前隐藏元素 / 涂盖打码、给静态截图加注释（红框 / 箭头 / 标签）",
-        "高亮某文件的 L5-L20 代码行并截图",
-        "移动端 / 设备模拟截图、带 cookie / 登录态截图"
+        "Web page screenshot, screen capture, full-page screenshot, long screenshot",
+        "Capture a specific element / region, partial screenshot, focus on an area",
+        "Hide or mask elements before capturing, annotate a still screenshot (highlight box / arrow / label)",
+        "Highlight lines L5-L20 of a file and screenshot it",
+        "Mobile / device-emulated screenshot, capture with cookies / a logged-in session"
       ],
       "entry": {
         "type": "python",

package/dist/runner.js

@@ -15,6 +15,7 @@ import { findSkill, SKILLS_DIR } from './registry.js';
 import { HANDLERS } from './handlers/index.js';
 import { EXIT, SkillError } from './errors.js';
 import { authChildEnv, ensureAuth } from './auth/ensure.js';
+import { emitBillingFooter } from './billing.js';
 /** Read the token override accepted by both the TS handlers and the Python skills. */
 function tokenFlag(args) {
     const value = args.token ?? args.priv_token;
@@ -34,6 +35,8 @@ export async function runSkill(skillName, opts) {
         });
         switch (skill.entry.type) {
             case 'python':
+                // Python children talk to ab-api themselves and print their own billing
+                // footer; nothing was charged through this process.
                 return await runPython(skill, opts.rawArgs, auth);
             case 'http':
             case 'builtin':
@@ -48,6 +51,12 @@ export async function runSkill(skillName, opts) {
         process.stderr.write(`❌ unexpected error: ${err.message}\n`);
         return EXIT.ERROR;
     }
+    finally {
+        // In `finally` because a run can fail *after* a billed step (e.g. the image
+        // generated and was charged, then the upload timed out) — spend gets
+        // reported either way. No-ops when nothing was charged.
+        emitBillingFooter();
+    }
 }
 async function runPython(skill, rawArgs, auth) {
     if (!skill.scriptAbsolutePath) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@remixmate/cli",
-  "version": "0.9.14",
+  "version": "0.9.16",
   "description": "AI media generation skills for Claude Code / Codex — 12 skills covering image, video, voice, digital human, web screenshot, web recording, script, template registry, rendering, Jianying export, and video deconstruction.",
   "type": "module",
   "main": "dist/index.js",

package/skills/export-jianying/version.json

@@ -3,5 +3,5 @@
   "repoName": "agent-skill-media-maker",
   "skillId": "554",
   "version": "V3",
-  "skillDescription": "剪映草稿生成技能，将素材URL打包为剪映可导入的草稿ZIP，支持从 RenderPlan 自动转换（调用 ab-api /file/generateJianYing）。\n\n当用户提到以下任何需求时，立即使用本 skill：\n- 导出剪映、剪映草稿、打包剪映、导入剪映\n- 将素材导出为剪映格式、生成剪映工程\n- 把视频/图片/音频打包成剪映草稿\n- 从 RenderPlan 导出剪映草稿\n\n即使用户没有明确说「剪映」，只要他们想要将素材打包为可在剪映中编辑的草稿格式，也要使用本 skill。"
+  "skillDescription": "Jianying (CapCut) draft-generation skill. Packages asset URLs into a draft ZIP that Jianying can import; supports automatic conversion from a RenderPlan (calls ab-api /file/generateJianYing).\n\nUse this skill as soon as the user mentions any of these intents:\n- Export to Jianying, Jianying draft, package for Jianying, import into Jianying\n- Export materials to the Jianying format, generate a Jianying project\n- Bundle video / image / audio into a Jianying draft\n- Export a Jianying draft from a RenderPlan\n\nEven when the user does not say \"Jianying\" explicitly, use this skill whenever they want to package materials into a draft that can be edited in Jianying."
 }

package/skills/gen-digital-human/SKILL.md

@@ -175,6 +175,18 @@ remixmate gen-digital-human --check-status --generation-id 123
 - Keep individual jobs under ~500 characters.
 - Tone and style of the script affect the perceived voice.
 
+## Credits
+
+Every run charges credits. The CLI prints a footer on stdout when it does:
+
+```
+💳 Charged 31 credits · balance 1,240
+```
+
+Relay it to the user whenever it appears — it is the only signal they get about what a
+generation cost, and the balance is the only warning before a run fails with
+`insufficient_credits`. Do not drop it from your summary.
+
 ## Error handling
 
 - **401** / **token missing** (non-OpenClaw): set `PRIV_TOKEN`.

package/skills/gen-digital-human/skill.json

@@ -12,13 +12,19 @@
     "type": "object",
     "properties": {
       "list_avatars": { "type": "boolean", "description": "List available digital-human avatars" },
-      "source": { "type": "string", "enum": ["jimeng", "hifly"], "description": "Filter by source" },
+      "mine": { "type": "boolean", "description": "With list_avatars=true: list the caller's own custom avatars instead of the public catalog. Custom avatars do not appear without this." },
+      "name": { "type": "string", "description": "With list_avatars=true: fuzzy-filter avatars by name" },
+      "source": { "type": "string", "enum": ["jimeng", "hifly"], "description": "Provider: jimeng is image-driven, hifly is video-driven. Usually inferred from the avatar; pass it explicitly when the avatar declares no source." },
       "gender": { "type": "string", "enum": ["male", "female"], "description": "Filter by gender" },
       "avatar_id": { "type": "number", "description": "Avatar id" },
       "text": { "type": "string", "description": "Narration text (TTS mode)" },
       "audio_url": { "type": "string", "description": "Audio URL (audio-driven mode)" },
-      "voice_id": { "type": "string", "description": "Voice id" },
-      "aspect_ratio": { "type": "string", "description": "Aspect ratio" },
+      "voice_id": { "type": "string", "description": "Voice id (TTS mode). Shares the Minimax catalog with gen-voice — call gen_voice with list_voices=true to see available ids rather than inventing one." },
+      "voice_name": { "type": "string", "description": "Voice display name, recorded alongside voice_id for bookkeeping. Does not affect synthesis." },
+      "aspect_ratio": { "type": "string", "description": "Aspect ratio: 9:16 / 16:9 / 3:4 / 1:1. Defaults to the avatar's own ratio." },
+      "prompt": { "type": "string", "description": "Action prompt describing how the avatar should perform, e.g. 'more hand gestures'" },
+      "check_status": { "type": "boolean", "description": "Status-check mode: poll an earlier job instead of starting a new one. Requires generation_id. Use this when a generate call timed out." },
+      "generation_id": { "type": "number", "description": "Job id to poll (required when check_status=true)" },
       "json_output": { "type": "boolean", "description": "Emit a JSON result (generate: { url, generationId }; list: { avatars }; check-status: { status, url }) instead of human-readable output" }
     },
     "required": []

package/skills/gen-digital-human/version.json

@@ -3,5 +3,5 @@
   "repoName": "agent-skill-media-maker",
   "skillId": "336",
   "version": "V8",
-  "skillDescription": "数字人口播视频技能，支持查询形象、TTS 口播、音频驱动口播和查询生成状态（调用 ab-api 数字人接口，即梦 / 飞影）。\n\n当用户提到以下任何需求时，立即使用本 skill：\n- 数字人、数字人视频、数字人口播、生成数字人视频\n- 用户提供已有音频 URL（MP3 等）、用外链音频驱动数字人口型\n- AI 主播、虚拟主播、虚拟人物视频、口播视频\n- 让数字人说话、让虚拟人说一段话、让 AI 人物播报\n- 查看/列出数字人形象、有哪些数字人可以用\n- 使用即梦/飞影数字人\n\n即使用户没有明确说「使用 AI」，只要他们想要让一个虚拟人物朗读/播报一段文字并生成视频，也要使用本 skill。"
+  "skillDescription": "Digital-human (talking-head) skill: list available avatars, run TTS-based or audio-driven lip-sync, and check the status of pending jobs. Backed by ab-api's digital-human endpoints (Jimeng / HiFly providers).\n\nUse this skill immediately whenever the user asks for any of:\n- Digital human, talking-head video, AI presenter, virtual host\n- \"Make a talking-head video out of this script / this audio file\"\n- List or browse digital-human avatars\n- Use Jimeng or HiFly to drive an avatar\n\nEven without an explicit \"use AI\", any request that turns text or audio into a synthesized presenter video should route here."
 }

package/skills/gen-image/SKILL.md

@@ -23,21 +23,28 @@ Wraps ab-api's `POST /model/genImg` (the same endpoint the web studio uses), aut
 
 ## Models and sizes
 
-Aligned with the gen-image handler's model presets and the backend `ModelGenImgDTO`:
-
-| LiteLLM `model` | Display name | Provider |
-|-----------------|--------------|----------|
-| `doubao/doubao-seedream-4-5-251128` | Seedream 4.5 | Volcano |
-| `doubao/doubao-seedream-5-0-260128` | Seedream 5.0 Lite | Volcano |
-| `gemini-3-pro-image` | Gemini 3 Pro | Google |
-| `gemini-3.1-flash-image-preview` | Gemini 3.1 Flash | Google |
-
-- **Seedream**: `--size` is an aspect ratio (e.g. `1:1`, `9:16`) or `WxH`; the backend may auto-upscale the 4.5 model to meet a minimum pixel count.
-- **Gemini**: `--size` is a backend-allowed aspect ratio (e.g. `1:1`, `16:9`); add `--resolution`: `1K` / `2K` / `4K` (default `1K`).
+Pass `--model` a short name. The authoritative roster — ids, aliases, per-model limits —
+lives in the backend catalog (`/model/capabilities`), which the CLI fetches at runtime;
+the names below are the stable aliases to use.
+
+| `--model` | What it is | Reference images |
+|-----------|------------|------------------|
+| `seedream` | Default. General-purpose, highest output resolution. | up to 14 |
+| `seedream-pro` | High-fidelity variant: better placement/element control, more faithful text rendering. Costs more per image. | up to 10 |
+| `gemini` | Gemini 3 Pro. | up to 4 |
+
+- **Seedream**: `--size` is an aspect ratio (e.g. `1:1`, `9:16`) or `WxH`. The backend maps the
+  ratio to that model's own pixel preset and rescales out-of-range sizes, so prefer a ratio
+  over explicit pixels.
+- **`seedream-pro`** additionally supports `3:2` / `2:3` / `21:9`, and caps output at ~2K
+  (about 4.6 MP). Asking it for 4K pixels gets scaled down, not rejected — use `seedream`
+  when you need a genuinely larger image.
+- **Gemini**: `--size` is a backend-allowed aspect ratio (e.g. `1:1`, `16:9`); add
+  `--resolution`: `1K` / `2K` / `4K` (default `1K`). `--resolution` is ignored by Seedream.
 
 ## Auth & environment
 
-No skill-local env file — the executing process inherits the system environment. Examples say `python`; on macOS you may need `python3`.
+No skill-local env file — the executing process inherits the system environment.
 
 - **Enterprise OpenClaw**: auth is already injected, **no need** for `PRIV_TOKEN` / `--priv-token`.
 - **Other environments**: configure the token. Without a token, non-interactive runs fail; interactive ones may prompt.
@@ -45,7 +52,7 @@ No skill-local env file — the executing process inherits the system environmen
 | Env var | Description | Default |
 |---------|-------------|---------|
 | `PRIV_TOKEN` | Tianyan token; `--priv-token` overrides | none |
-| `MM_IMAGE_MODEL` | Default model id | `doubao/doubao-seedream-4-5-251128` |
+| `MM_IMAGE_MODEL` | Default model, as a `--model` value | backend default (`seedream`) |
 | `MM_API_BASE_URL` | API root; `--api-base-url` overrides | `https://api.remixmate.com/api` |
 | `AGENT_NAME` | Optional `x-invoke-agent` header | none |
 
@@ -67,17 +74,30 @@ remixmate gen-image \
 ```bash
 remixmate gen-image \
   --prompt "<image description>" \
-  --model gemini-3-pro-image \
+  --model gemini \
   --size "16:9" \
   --resolution "2K"
 ```
 
+```bash
+# High-fidelity: precise placement, legible on-image text
+remixmate gen-image \
+  --prompt "<image description>" \
+  --model seedream-pro \
+  --size "16:9"
+```
+
 ### Image-to-image (reference image)
 
 Reference images accept local file paths, HTTPS URLs, or data URIs. Pass `--reference` multiple times for multiple references.
 
-- **Seedream**: up to **14** reference images, `--image-strength` controls reference influence.
-- **Gemini**: up to **4** reference images.
+- **`seedream`**: up to **14** reference images, `--image-strength` controls reference influence.
+- **`seedream-pro`**: up to **10** reference images. Best choice when the edit has to land in a
+  specific spot — describe the target region in the prompt (e.g. "in the marked area at the
+  bottom left") and it holds position far better than `seedream`.
+- **`gemini`**: up to **4** reference images.
+
+Over-the-limit runs fail fast in the CLI, before spending credits.
 
 ```bash
 # URL reference
@@ -109,7 +129,7 @@ remixmate gen-image \
 | Flag | Description | Default |
 |------|-------------|---------|
 | `-p` / `--prompt` | Description (required) | — |
-| `-m` / `--model` | Model id | see `MM_IMAGE_MODEL` |
+| `-m` / `--model` | `seedream` / `seedream-pro` / `gemini` | see `MM_IMAGE_MODEL` |
 | `-s` / `--size` | Seedream: ratio or WxH; Gemini: ratio | `1:1` |
 | `--resolution` | Gemini only: `1K` / `2K` / `4K` | `1K` |
 | `-n` | Number of images, 1–4 | `1` |
@@ -122,6 +142,20 @@ remixmate gen-image \
 | `--api-base-url` | Override API root | see above |
 | `--priv-token` | Override token | see above |
 
+## Credits
+
+Every run charges credits, per image and **per model** — `seedream-pro` costs noticeably more
+per image than `seedream`, so don't reach for it by default. The CLI prints a footer on stdout
+when it charges:
+
+```
+💳 Charged 31 credits · balance 1,240
+```
+
+Relay it to the user whenever it appears — it is the only signal they get about what a
+generation cost, and the balance is the only warning before a run fails with
+`insufficient_credits`. Do not drop it from your summary.
+
 ## Error handling
 
 - **401** / **token missing** (non-OpenClaw): set `PRIV_TOKEN`.

package/skills/gen-image/skill.json

@@ -4,7 +4,7 @@
   "tier": "atomic",
   "category": "asset",
   "title": "AI Image Generation",
-  "description": "AI image generation: produce an image from a text prompt. Supports Seedream and Gemini models, plus image-to-image with reference images.",
+  "description": "AI image generation: produce an image from a text prompt. Supports the Seedream family (including a high-fidelity 'pro' variant) and Gemini, plus image-to-image with reference images.",
   "auth": "required",
   "envVars": ["PRIV_TOKEN", "MM_API_BASE_URL", "AGENT_NAME", "MM_IMAGE_MODEL"],
   "entry": { "type": "http", "handler": "gen-image" },
@@ -12,12 +12,20 @@
     "type": "object",
     "properties": {
       "prompt": { "type": "string", "description": "Image description (required)" },
-      "model": { "type": "string", "description": "Model id" },
+      "model": { "type": "string", "description": "Model: 'seedream' (default), 'seedream-pro' (high fidelity, precise placement and on-image text; costs more per image), or 'gemini'" },
       "size": { "type": "string", "description": "Aspect ratio or WxH, e.g. 1:1, 9:16" },
       "resolution": { "type": "string", "enum": ["1K", "2K", "4K"], "description": "Output resolution (Gemini only)" },
       "n": { "type": "number", "description": "Number of images, 1-4" },
-      "reference": { "type": "string", "description": "Reference image path or URL" },
+      "reference": {
+        "type": "array",
+        "items": { "type": "string" },
+        "description": "Reference images for image-to-image: local file path, https URL, or data URI. Pass multiple to blend several references (seedream: max 14, seedream-pro: max 10, gemini: max 4 — over the limit fails before spending credits)."
+      },
+      "image_strength": { "type": "number", "description": "How strongly the reference images influence the result, 0-1 (Seedream family only). Omit to use the backend default." },
+      "guidance_scale": { "type": "number", "description": "Prompt-adherence strength, where supported. Omit to use the backend default." },
       "negative_prompt": { "type": "string", "description": "Negative prompt — content to avoid" },
+      "seed": { "type": "number", "description": "Random seed. Pass the same seed with the same prompt and model to make a run reproducible." },
+      "watermark": { "type": "boolean", "description": "Add a watermark to the output. Only true has an effect; there is no opt-out override of the backend default." },
       "json_output": { "type": "boolean", "description": "Emit a JSON result ({ urls: [...] }) instead of human-readable output" }
     },
     "required": ["prompt"]

package/skills/gen-image/version.json

@@ -3,5 +3,5 @@
   "repoName": "agent-skill-media-maker",
   "skillId": "337",
   "version": "V9",
-  "skillDescription": "AI 生图技能，根据文字描述生成图片，也支持参考图进行图生图（调用 ab-api /model/genImg，支持 Seedream 与 Gemini）。\n\n当用户提到以下任何需求时，立即使用本 skill：\n- AI 生图、AI 画图、文生图、文字生成图片、生成图像、帮我画、生成一张图\n- 图生图、以图生图、参考图、风格迁移、图片变体\n- 使用 doubao / 豆包 / seedream、Gemini 等生成图片\n- 用户提供图片提示词并希望生成图片\n\n即使用户没有明确说「使用 AI」，只要他们想要根据描述生成图片，也要使用本 skill。"
+  "skillDescription": "AI image generation skill: produce an image from a text prompt, or do image-to-image with reference images. Backed by ab-api's `/model/genImg` (Seedream and Gemini families).\n\nUse this skill immediately whenever the user asks for any of:\n- AI image generation, text-to-image, \"draw me ...\", \"generate an image of ...\"\n- Image-to-image, reference image, style transfer, image variation\n- Generate an image with Doubao / Seedream / Gemini\n- Provide a prompt and ask for an image\n\nEven without an explicit \"use AI\", any request that turns a description into an image should route here."
 }

package/skills/gen-script/SKILL.md

@@ -293,26 +293,28 @@ Apply the following principles when producing the DSL:
    "text": "Alibaba's open-source AI video-editing project\nalready has 4.3k stars on GitHub"
    ```
    instead of one continuous paragraph.
-4. **Moderate scene count**: 30-second videos work well with 4–6 scenes, 60-second videos with 6–10.
-5. **Leave room for templates**: pick generic layouts; do not assume a specific template implementation.
-6. **Image model allowlist**: every `type: image` + `source: gen-image` `AssetRef`'s `payload.model` **must** be one of the values in the table below. **Never** use display names, short forms, or made-up ids (e.g. `seedream`, `gemini-flash`, etc.).
+5. **Moderate scene count**: 30-second videos work well with 4–6 scenes, 60-second videos with 6–10.
+6. **Leave room for templates**: pick generic layouts; do not assume a specific template implementation.
+7. **Image model allowlist**: every `type: image` + `source: gen-image` `AssetRef`'s `payload.model` **must** be one of the values in the table below. **Never** use display names, short forms, or made-up ids (e.g. `seedream`, `gemini-flash`, etc.).
 
 ### Allowlist `model` values aligned with gen-image
 
-Mirrors the `gen-image` skill and `GEN_IMG_MODEL_PRESETS` in `gen_image.py`. **Only** the following four values are allowed (copy verbatim, including prefix and version):
+The roster is owned by the backend catalog (`/model/capabilities`), which `gen-image` resolves at
+runtime. A DSL is persisted and replayed later, so write the **full id** — copy verbatim, including
+prefix and version — not a short alias:
 
-| LiteLLM `model` | Display name | Provider |
-|-----------------|--------------|----------|
-| `doubao/doubao-seedream-4-5-251128` | Seedream 4.5 | Volcano |
-| `doubao/doubao-seedream-5-0-260128` | Seedream 5.0 Lite | Volcano |
-| `gemini-3-pro-image` | Gemini 3 Pro | Google |
-| `gemini-3.1-flash-image-preview` | Gemini 3.1 Flash | Google |
+| `payload.model` | Display name | Provider | Notes |
+|-----------------|--------------|----------|-------|
+| `doubao/doubao-seedream-5-0-260128` | Seedream 5.0 Lite | Volcano | Default. Highest output resolution, up to 14 reference images. |
+| `doubao/doubao-seedream-5-0-pro-260628` | Seedream 5.0 Pro | Volcano | High fidelity: precise element placement, faithful on-image text. Up to 10 reference images, caps out around 2K. Costs noticeably more per image. |
+| `gemini-3-pro` | Gemini 3 Pro | Google | Up to 4 reference images. |
 
 **Agent behavior (avoid accidentally rewriting `model`)**:
 
-- `gen_script.py` already writes a valid `payload.model` (currently `gemini-3.1-flash-image-preview` by default). When the user only asks to refine narration, change `payload.prompt`, add or remove scenes, etc. and does **not** ask to change the image model, the agent **must keep** each image asset's original `payload.model` — do not replace it under the guise of "polishing the script".
-- **Only when the user explicitly asks to change the image model** (e.g. switches to Seedream or a different Gemini), update the corresponding image `AssetRef`'s `payload.model` to the matching row id from the table. Writing a display name into JSON is wrong.
-- When creating a new image `AssetRef`, pick one of the values above for `payload.model`; default to `gemini-3.1-flash-image-preview` to match the script, or to whichever value the user specified.
+- `gen_script.py` already writes a valid `payload.model` (`doubao/doubao-seedream-5-0-260128` unless `DEFAULT_IMAGE_MODEL` overrides it). When the user only asks to refine narration, change `payload.prompt`, add or remove scenes, etc. and does **not** ask to change the image model, the agent **must keep** each image asset's original `payload.model` — do not replace it under the guise of "polishing the script".
+- **Only when the user explicitly asks to change the image model** (e.g. switches to the Pro variant or to Gemini), update the corresponding image `AssetRef`'s `payload.model` to the matching row id from the table. Writing a display name or an alias into JSON is wrong.
+- When creating a new image `AssetRef`, pick one of the values above for `payload.model`; default to `doubao/doubao-seedream-5-0-260128` to match the script, or to whichever value the user specified.
+- If a run fails with an unknown-model error, the catalog has moved on from this table — check `/model/capabilities` rather than guessing a version string.
 
 ## Error handling
 
@@ -325,11 +327,5 @@ Mirrors the `gen-image` skill and `GEN_IMG_MODEL_PRESETS` in `gen_image.py`. **O
 | File | Purpose |
 |------|---------|
 | `gen_script.py` | Core script — produces the Video DSL JSON from a topic. |
-| `dsl.json` | DSL example produced from the generic template (debug reference). |
-| `script.json` | Intermediate script-generation result (example). |
-| `html_slide.json` | DSL example for the html-slide template. |
-| `knowledge_card_dsl.json` | DSL example for the knowledge-card scene. |
-| `knowledge_card_script.json` | Intermediate knowledge-card script result. |
-| `script_knowledge_card.json` | Full knowledge-card script example. |
-
-These JSON files are reference data for development / debugging; they do not participate in runtime logic. The authoritative DSL examples live under `template-registry/video_dsl/schema/examples/`.
+
+The authoritative DSL examples live under `template-registry/video_dsl/schema/examples/`; list them with `template_registry list_examples=true`.

package/skills/gen-script/version.json

@@ -3,5 +3,5 @@
   "repoName": "agent-skill-media-maker",
   "skillId": "474",
   "version": "V11",
-  "skillDescription": "视频脚本生成技能，将用户主题转化为结构化 Video DSL（JSON），描述视频的完整结构、素材需求与叙事逻辑。\n\n当用户提到以下任何需求时，立即使用本 skill：\n- 帮我写视频脚本、生成视频脚本、视频策划、写分镜脚本\n- 做一个短视频、帮我规划视频内容、生成视频 DSL\n- 把主题转成视频结构、视频内容规划\n\n即使用户没有明确说「生成 DSL」，只要他们想要把一个主题变成视频内容结构，也要使用本 skill。"
+  "skillDescription": "Video-script generation skill. Turns a user-supplied topic into a structured Video DSL (JSON) that describes the full video — scene structure, asset requirements, and narrative flow.\n\nUse this skill as soon as the user mentions any of these intents:\n- Write a video script, generate a video script, plan a video, write storyboards\n- Create a short video, plan video content, generate a Video DSL\n- Turn a topic into a video structure / video content plan\n\nEven when the user does not say \"generate the DSL\", use this skill whenever they want to turn a topic into a structured video plan.\n\n⚠️ Stop-and-confirm gate: after this skill returns a DSL, show the full script and wait for the user's explicit confirmation. Never call `prepare_video_assets` in the same turn."
 }