← 返回 Skills 市场
cruciata

social-auto-tool-builder-1.1.0

作者 范士轶 · GitHub ↗ · v1.1.0 · MIT-0
cross-platform ✓ 安全检测通过
760
总下载
2
收藏
1
当前安装
2
版本数
在 OpenClaw 中安装
/install social-auto-tool-builder
功能描述
复用“小红书自动回复项目”实战经验,快速构建新的本地AI自动化工具(含多平台选择器映射模板)
使用说明 (SKILL.md)

\r \r

社媒自动化工具构建 Skill v1.1.0\r

\r

目标\r

把“新自动化工具”的制作过程标准化,并在 v1.0.0 基础上新增多平台适配能力:\r

  • Python + Playwright(persistent_context)\r
  • 本地 Ollama(意图识别 + 回复生成)\r
  • 规则过滤 + AI 生成混合策略\r
  • 先 dry-run 再真实发送\r
  • 参数化运行 + Windows EXE 打包\r
  • 小红书 / 抖音 / 快手选择器映射模板\r \r ---\r \r

先决输入(必须确认)\r

  1. 目标平台(xiaohongshu / douyin / kuaishou)\r
  2. 入口 URL(通知页/消息页)\r
  3. 已回复判定规则(例如“作者”)\r
  4. 发送成功信号(toast 或状态文本)\r
  5. 时间窗口与轮询间隔\r
  6. 单轮最多处理条数\r \r 默认值(信息不全时):\r
  • interval-minutes=5\r
  • recent-hours=1\r
  • max-replies=0\r
  • --once dry-run\r \r ---\r \r

多平台选择器映射模板(核心)\r

\r 在主程序中维护一个映射表(先填模板,后按真实 DOM 覆盖):\r \r

SELECTOR_MAP = {\r
    "xiaohongshu": {\r
        "url": "https://www.xiaohongshu.com/notification",\r
        "comment_item": ".interaction-item, .comment-item",\r
        "comment_text": ".interaction-content, .comment-text, .content",\r
        "reply_trigger": "button:has(svg.reply-icon), svg.reply-icon, .reply-btn",\r
        "input": "p#content-textarea.content-input, [contenteditable='true'].content-input, textarea.comment-input",\r
        "submit": "button.submit, button:has-text('发送')",\r
        "success": "text=评论成功",\r
        "replied_marker": "作者"\r
    },\r
    "douyin": {\r
        "url": "\x3CTO_FILL_DOUYIN_URL>",\r
        "comment_item": "\x3CTO_FILL>",\r
        "comment_text": "\x3CTO_FILL>",\r
        "reply_trigger": "\x3CTO_FILL>",\r
        "input": "\x3CTO_FILL>",\r
        "submit": "\x3CTO_FILL>",\r
        "success": "\x3CTO_FILL_SUCCESS_TEXT>",\r
        "replied_marker": "\x3CTO_FILL_MARKER>"\r
    },\r
    "kuaishou": {\r
        "url": "\x3CTO_FILL_KUAISHOU_URL>",\r
        "comment_item": "\x3CTO_FILL>",\r
        "comment_text": "\x3CTO_FILL>",\r
        "reply_trigger": "\x3CTO_FILL>",\r
        "input": "\x3CTO_FILL>",\r
        "submit": "\x3CTO_FILL>",\r
        "success": "\x3CTO_FILL_SUCCESS_TEXT>",\r
        "replied_marker": "\x3CTO_FILL_MARKER>"\r
    }\r
}\r
```\r
\r
适配规则:\r
- 先读平台映射,再执行统一流程。\r
- 每个关键 selector 至少保留 1 个 fallback。\r
- 任何平台都先跑 dry-run 列候选,再启用真实发送。\r
\r
---\r
\r
## 标准流程(6 Phase)\r
\r
### Phase 1:骨架搭建\r
- 主程序拆分:AI 客户端、自动化模块、参数解析。\r
- 支持参数:\r
  - `--platform`\r
  - `--interval-minutes`\r
  - `--recent-hours`\r
  - `--max-replies`\r
  - `--once`\r
  - `--interactive`\r
\r
### Phase 2:平台映射接入\r
- 将平台 selector 映射注入 `Automation`。\r
- URL、评论列表、输入框、发送按钮、成功信号全部来自映射。\r
\r
### Phase 3:真实 DOM 校准\r
- 让用户提供截图/DOM 片段,增量修正映射。\r
- 每次只改一个动作链路(打开弹窗、输入、发送)。\r
\r
### Phase 4:安全与准确\r
- 已回复过滤(marker)\r
- 时间窗口过滤(recent-hours)\r
- 去重(`(text, age)`)\r
- 2-5 秒随机延迟\r
\r
### Phase 5:验证\r
- dry-run:列出候选不发送\r
- 用户确认后真实发送\r
- 必须等待成功信号\r
\r
### Phase 6:交付\r
- EXE 打包(含 Playwright 浏览器路径兼容)\r
- QUICK_START + FAQ\r
- Git 提交与推送\r
\r
---\r
\r
## 质量闸门(必须通过)\r
1. `python -m py_compile auto_responder_production.py`\r
2. `--once` 模式可运行\r
3. dry-run 候选与人工预期一致\r
4. 至少 1 条真实发送成功信号\r
5. EXE 参数模式可运行\r
6. EXE 交互模式可运行\r
\r
---\r
\r
## Windows 命令模板\r
\r
```powershell\r
# 安装依赖\r
pip install -r requirements.txt\r
python -m playwright install chromium\r
\r
# 本地模型检查\r
python -c "import requests;print(requests.get('http://127.0.0.1:11434/api/tags',timeout=5).status_code)"\r
\r
# 单轮 dry-run\r
python auto_responder_production.py --platform xiaohongshu --once --recent-hours 1 --max-replies 3\r
\r
# 构建 EXE\r
powershell -ExecutionPolicy Bypass -File .\build_exe.ps1\r
\r
# EXE 参数模式\r
.\dist\auto_responder.exe --platform xiaohongshu --interval-minutes 5 --recent-hours 2 --max-replies 3\r
\r
# EXE 交互模式\r
.\dist\auto_responder.exe --interactive\r
```\r
\r
---\r
\r
## 可复用 Prompt(新项目直接用)\r
\r
```text\r
按 social-auto-tool-builder v1.1.0 的流程,帮我做一个【平台名】自动化工具。\r
要求:\r
1) Python + Playwright + 本地 Ollama\r
2) 用 persistent_context 保存登录状态\r
3) 先 dry-run 列候选,再真实发送\r
4) 支持 --platform --interval-minutes --recent-hours --max-replies --once --interactive\r
5) 用多平台选择器映射模板实现,允许按我提供的DOM增量修正\r
6) 最后打包EXE并给 QUICK_START\r
```\r
\r
---\r
\r
## 输出物检查清单\r
- [ ] `skill.yaml` 与 `SKILL.md` 完整\r
- [ ] 选择器映射模板已建立\r
- [ ] 平台参数化设计明确\r
- [ ] 验证与交付步骤可执行\r
安全使用建议
This skill appears coherent with its stated goal of producing a local Playwright + Ollama auto-responder. Before using or following its instructions: 1) Understand that the resulting tool will automate actions as your logged-in browser user — test carefully in dry-run mode and on non-production accounts to avoid policy/ToS violations. 2) The skill assumes you run a local Ollama server (127.0.0.1:11434) and have Playwright/browser binaries installed; ensure those are legitimate and up-to-date. 3) The skill itself asks for no secrets, but the tools you build may require login cookies or credentials stored in the browser profile — keep those local and secure. 4) When packaging an EXE, verify the build scripts and any bundled files yourself to avoid accidentally including sensitive data. If you want a higher-assurance review, provide the actual automation code (auto_responder_production.py, build_exe.ps1, requirements.txt) so those files can be inspected for unsafe behavior (remote endpoints, credential capture, obfuscated operations).
功能分析
Type: OpenClaw Skill Name: social-auto-tool-builder Version: 1.1.0 The skill bundle provides a legitimate framework for building social media automation tools using Python, Playwright, and local Ollama. The instructions in SKILL.md and skill.yaml outline a standard development lifecycle, including dry-run modes, selector mapping for platforms like Xiaohongshu and Douyin, and EXE packaging. No evidence of data exfiltration, malicious persistence, or harmful prompt injection was found; the tool focuses on local execution and user-controlled automation.
能力评估
Purpose & Capability
Name/description, SKILL.md, and skill.yaml all describe building a local AI + Playwright auto-responder and the instructions only reference Playwright, local Ollama, packaging to EXE, selector templates and dry-run vs send flows — these are coherent and expected for that purpose.
Instruction Scope
SKILL.md explicitly instructs reading DOM/snippets/screenshots to calibrate selectors, running dry-runs, and then performing sends; it does not instruct the agent to read unrelated system files or exfiltrate secrets. It does reference checking a local Ollama HTTP endpoint (127.0.0.1) and using persistent_context for browser login state, which is appropriate for the stated functionality.
Install Mechanism
This is an instruction-only skill with no install spec and no code files. The SKILL.md suggests using common package installs (pip, playwright install) and EXE packaging tools; nothing is pulled from arbitrary URLs in the skill itself.
Credentials
The skill declares no required environment variables, credentials, or config paths. The only runtime network reference is to localhost:11434 (local Ollama), which matches the stated use of a local model. No unrelated tokens or secrets are requested.
Persistence & Privilege
Skill flags are default (not always:true). It does not request persistent platform-wide privileges or instruct modification of other skills or agent-wide configs. It does suggest storing browser login state (persistent_context) which is expected for an automation tool that must act as a logged-in user.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install social-auto-tool-builder
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /social-auto-tool-builder 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.1.0
social-auto-tool-builder v1.1.0 introduces multi-platform selector mapping for broader automation support. - Added multi-platform selector mapping template for Xiaohongshu, Douyin, and Kuaishou. - CLI and workflow now require specifying target platform with --platform. - Main workflow and EXE support platform-agnostic operation via selector table. - Updated standard input/checklist to ensure platform-parametric adaptation and delivery. - Documentation, prompts, and Windows command templates revised to reflect multi-platform design.
v1.0.0
小红书自动化工具构建流程
元数据
Slug social-auto-tool-builder
版本 1.1.0
许可证 MIT-0
累计安装 1
当前安装数 1
历史版本数 2
常见问题

social-auto-tool-builder-1.1.0 是什么?

复用“小红书自动回复项目”实战经验,快速构建新的本地AI自动化工具(含多平台选择器映射模板). 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 760 次。

如何安装 social-auto-tool-builder-1.1.0?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install social-auto-tool-builder」即可一键安装,无需额外配置。

social-auto-tool-builder-1.1.0 是免费的吗?

是的,social-auto-tool-builder-1.1.0 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

social-auto-tool-builder-1.1.0 支持哪些平台?

social-auto-tool-builder-1.1.0 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 social-auto-tool-builder-1.1.0?

由 范士轶(@cruciata)开发并维护,当前版本 v1.1.0。

💬 留言讨论