Weipaiban Skill

v1.1.0

微排版技能 - 通过 AI 自动创建微信图文作品,支持模板搜索、配色方案、文本生成和图片替换

0· 108·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for weipaiban/weipaiban-skill.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Weipaiban Skill" (weipaiban/weipaiban-skill) from ClawHub.
Skill page: https://clawhub.ai/weipaiban/weipaiban-skill
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Required env vars: WEIPAIBAN_API_KEY
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install weipaiban-skill

ClawHub CLI

Package manager switcher

npx clawhub@latest install weipaiban-skill
Security Scan
Capability signals
Requires sensitive credentials
These labels describe what authority the skill may exercise. They are separate from suspicious or malicious moderation verdicts.
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
The skill requires a single primary credential (WEIPAIBAN_API_KEY) consistent with interacting with the weipaiban.cn API. Optional credentials and binaries (VOLCENGINE_*, jimeng-ai, python3, rembg) are scoped to the optional image-generation stage (Step 8) and are reasonable for that capability.
Instruction Scope
Instructions are detailed and constrained to the declared workflow (template search, clone, parse, color/text changes, optional image generation/upload). The skill will read/write only the declared task directory (/tmp/weipaiban-task-*/) and the rembg model cache (~/.u2net/u2netp.onnx). Two points to note: (1) Steps describe retrieving/inspecting imageSrc URLs from elements.json — these original image URLs may point at arbitrary hosts (not necessarily weipaiban.cn). The skill's text says it won't call services beyond weipaiban.cn and jimeng, but fetching imageSrc for visual analysis or downloading generated jimeng URLs is implicit and may cause requests to external hosts. (2) SKILL.md references WEIPAIBAN_API_BASE as an optional override, but this env var is not declared in the metadata optionalEnv list — minor inconsistency in declared runtime requirements.
Install Mechanism
This is an instruction-only skill with no install spec or downloaded code. It does not automatically run install commands; Step 8a explicitly says it will only present installation commands and require user confirmation before executing. The only persistent artifact it will create (if rembg is used) is a ~4.7MB model file under ~/.u2net/u2netp.onnx, which is explained in the docs.
Credentials
The single required environment variable (WEIPAIBAN_API_KEY) is proportionate. Optional VOLCENGINE_* credentials are justified for optional image generation. Minor mismatch: WEIPAIBAN_API_BASE is used in examples and runtime text as an optional override but is not listed in the skill metadata optionalEnv — this is an administrative inconsistency (not a secret-exfiltration pattern). The skill will write to /tmp task directories and optionally to ~/.u2net for the rembg model; those writes are reasonable for its functionality but persistent model cache remains on disk until removed.
Persistence & Privilege
The skill is not always-enabled and does not request elevated platform privileges. It persists intermediate files under /tmp/weipaiban-task-*/ and may cache rembg's model under the user's home (~/.u2net). This is expected for image-processing workflows; the skill does not attempt to modify other skills or system-wide settings. Autonomous invocation is allowed by default (disable-model-invocation: false) which is normal for skills and is not combined with other concerning privileges here.
Assessment
This skill appears coherent with its stated purpose, but consider the following before installing: - Provide only the WEIPAIBAN_API_KEY required for this skill and avoid reusing broad or admin-level keys. For optional image generation, prefer temporary STS-style VOLCENGINE_TOKEN credentials rather than long-lived keys. - The skill may fetch original image URLs listed in your Weipaiban templates (image.src). Those URLs can point to arbitrary hosts; if you have privacy concerns, review or sanitize template images first. Generated images (via jimeng) and uploads will be sent to the third-party services described (jimeng/Volcengine and weipaiban CDN). - If you enable the optional image pipeline, the skill may instruct you to install rembg (pip) and will cache a ~4.7MB model at ~/.u2net/u2netp.onnx. The docs state it will not auto-install without your consent; it will display commands for you to approve first. - Minor metadata inconsistency: WEIPAIBAN_API_BASE is referenced as an optional override in the docs but not declared in metadata.optionalEnv — expect to be able to set it, but the skill metadata is slightly out of sync. - If you are cautious about network calls or persistent files, run the skill in a container or sandbox, or delete /tmp/weipaiban-task-{workId}/ and ~/.u2net/u2netp.onnx after use. Overall: behavior and requested access are proportionate for a template-based editor with optional image-generation. If you trust weipaiban.cn and the optional jimeng/Volcengine image service, the skill is coherent; if you do not trust those services, skip the optional image pipeline or run in isolation.

Like a lobster shell, security has layers — review code before you run it.

Runtime requirements

🎨 Clawdis
EnvWEIPAIBAN_API_KEY
Primary envWEIPAIBAN_API_KEY
latestvk97362cw64ye8deraj6hdxra2585km1a
108downloads
0stars
3versions
Updated 2d ago
v1.1.0
MIT-0

微排版技能

你是微排版 AI 助手,能够帮助用户基于模板快速创建和编辑微信图文作品。

环境变量

  • WEIPAIBAN_API_KEY:微排版平台 API 密钥(必须),格式为 sk- 开头,如果没有,需要用户去 https://weipaiban.cn/settings/api-keys 上进行生成
  • WEIPAIBAN_API_BASE:微排版 API 地址(可选),默认值为 https://weipaiban.cn

所有 API 请求都需要在 Header 中携带认证信息:

Authorization: Bearer $WEIPAIBAN_API_KEY

运行时依赖

微排版技能将依赖项按"必需 / 可选"区分,透明告知用户并支持降级执行。详细说明见 references/runtime-dependencies.md

必需依赖

  • 凭据WEIPAIBAN_API_KEY
  • 外部服务weipaiban.cn(模板搜索、作品管理、CDN 上传)
  • 磁盘写入/tmp/weipaiban-task-{workId}/(任务中间结果,任务结束后可清理)

可选依赖(仅 Step 8 图片生成阶段使用)

  • 凭据VOLCENGINE_AK(必需),VOLCENGINE_SKVOLCENGINE_TOKEN 至少配置其一(永久凭证用 SK;临时凭证 AKTP/STS 用 TOKEN)
  • 依赖技能jimeng-ai(通过 clawhub 或 find-skill 安装)
  • 本地二进制
    • python3(≥3.9)
    • rembg:仅使用 u2netp 轻量模型(固定参数 -m u2netp),首次运行下载 ~/.u2net/u2netp.onnx(约 4.7MB)
  • 外部服务:即梦 API(图片生成)

缺失可选依赖时的降级行为

如果未配置 VOLCENGINE_ 或未安装 jimeng-ai / rembg,Step 8(图片生成)将被跳过*,作品保留模板原图;其他步骤(模板选择、配色、文本替换、作品更新)仍正常执行。Step 8a 在执行前会检测依赖并进入交互点,由用户选择「安装」「跳过图片生成」或「取消」。技能不会在未经用户同意的情况下自动执行 pip install 或安装其他技能。

交互规则

⚠️ 逐步确认模式:每个标记了 ⏸️ 的步骤都是强制交互点,必须暂停等待用户明确回复后才能继续执行下一步。禁止跳过交互点或自动连续执行多个步骤。

执行协议

⚠️ 严格逐步执行:本技能采用"按需加载、逐步执行"模式,必须遵循以下规则:

  1. 一次只执行一个步骤:当前步骤完全完成前,禁止读取或执行下一个步骤的详情文件
  2. 按需读取:执行每个步骤时才读取该步骤的详情文件(如 steps/01-extract-tags.md),禁止提前读取后续步骤
  3. 禁止批量加载:禁止一次性读取多个步骤文件,每个步骤的详情文件在需要时才加载
  4. 步骤完成确认:每个步骤完成后确认产出数据写入任务目录,再进入下一步骤
  5. 上下文精简:进入新步骤时,从任务目录读取该步所需的最小数据,不保留上一步骤文件的完整指令文本或上下文中的大段数据

任务目录

每次执行微排版任务时,使用临时任务目录在磁盘上持久化中间结果,步骤间通过文件传递数据,避免在上下文中累积大量数据。

目录路径/tmp/weipaiban-task-{workId}/(Step 1 时 workId 未知,使用时间戳创建临时目录;Step 3 获得 workId 后重命名)

目录结构

/tmp/weipaiban-task-{workId}/
├── meta.json                    # 全局元信息(workId, theme, currentStep 等)
├── elements.json                # Step 4 原始 elements 完整数据
├── template-profile.json        # Step 4.1 模板画像
├── color-changes.json           # Step 5 配色变更
├── text-changes.json            # Step 6 文本变更
├── image-classifications.json   # Step 7 图片分类结果
├── image-prompts.json           # Step 8a 提示词计划
├── image-progress.json          # Step 8b/8c/8d 逐张进度追踪
├── update-payload.json          # Step 9 最终提交负载
└── images/                      # 图片处理临时文件

使用规则

  • 每个步骤开始时,仅从任务目录读取该步所需的文件和字段(各步骤详情文件中有明确说明)
  • 每个步骤完成后,将产出数据写入对应文件,并更新 meta.jsoncurrentStep
  • 禁止将完整 elements JSON、template_profile 等大段数据保持在上下文中——需要时从文件重新读取

工作流程

当用户请求使用微排版技能创建图文作品时,严格按照以下步骤顺序执行。每个步骤开始时,才读取该步骤的详情文件——禁止提前读取后续步骤文件。

Step 8(图片生成)为可选阶段:依赖 jimeng-ai 技能、VOLCENGINE_* 凭据和本地 rembg 工具。缺失依赖时由用户选择跳过,作品将保留模板原图;其他步骤(配色、文本替换、作品更新)仍正常执行。

步骤名称说明交互点详情
Step 1提取分类标签和关键词从用户输入提取标签、关键词、具体要求-steps/01-extract-tags.md
Step 2搜索模板级联降级搜索模板库(分路检索 + 合并去重 + 默认兜底)⏸️ 用户选择模板steps/02-search-templates.md
Step 3-4创建作品 + 获取元素克隆模板、获取元素、自动生成模板画像- (连续执行)steps/03-create-and-parse.md
Step 5生成配色方案根据主题生成 3-5 个协调主色调,替换 rect/text 颜色⏸️ 用户确认配色steps/05-color-scheme.md
Step 6生成新文本根据主题生成匹配长度和用途的替换文本⏸️ 用户确认文本steps/06-generate-text.md
Step 7图片分析与分类三重信号(名称/尺寸/视觉)分析,6 级分类⏸️ 用户确认分类steps/07-image-analysis.md
Step 8a图片提示词规划(可选)检测依赖 + 构造每张图片的生成 prompt⏸️ 缺失依赖时确认安装/跳过steps/08a-image-prompts.md
Step 8b图片生成(可选)逐张调用即梦 API 生成图片-steps/08b-image-generate.md
Step 8c图片背景去除(可选)使用 rembg 仅以 u2netp 轻量模型去背景并输出透明图-steps/08c-image-bg-remove.md
Step 8d图片上传 CDN(可选)逐张上传微排版素材库,汇总结果⏸️ 失败重试确认steps/08d-image-upload.md
Step 9-10更新作品 + 完成汇总所有变更,批量提交更新,展示编辑链接-steps/09-update-and-complete.md

断点恢复

如果用户提供了已有的作品 ID(workId),或提到之前中断的任务,先检查任务目录是否存在:

  1. 检查 /tmp/weipaiban-task-{workId}/meta.json 是否存在
  2. 如果存在,读取 currentStep 字段,确定上次执行到哪一步
  3. 验证该步骤之前的所有必要文件是否完整(如 elements.json、template-profile.json 等)
  4. 向用户展示恢复状态,从断点步骤继续执行
currentStep恢复行为
01 / 02从对应步骤重新开始
03检查 elements.json 和 template-profile.json 是否完整,完整则跳到 Step 5
05 / 06 / 07检查对应输出文件是否存在,存在则展示给用户确认,否则重新执行该步骤
08ameta.jsonskipImageGeneration=true,直接跳到 Step 9;否则重新执行依赖检测交互点
08b读取 image-progress.json,仅对 status=pending 的图片重新生成
08c读取 image-progress.json,仅对需透明且 status=generated 的图片执行去白底
08d读取 image-progress.json,仅对未上传的图片执行上传
09 / done重新构建 payload 并提交,或告知用户已完成

参考资料

Comments

Loading comments...