← Back to Skills Marketplace
342
Downloads
0
Stars
1
Active Installs
1
Versions
Install in OpenClaw
/install inkroam-topic-expert
Description
选题专家：全网热点监控 → 筛选打分 → 初步三要素 → 写入飞书选题库 → 推送Telegram。Use when user asks to 启动选题专家 / 获取选题推荐 / 查看热点 / 帮我选题。
README (SKILL.md)
选题专家 Skill
内容生产流水线的上游，负责"找什么值得写"。
【选题专家】→ 飞书选题库 → 审批 → 笔尖专家 → 发布专家
配置
环境变量
在工作区创建 .env 或在 config.json 中配置：
export FEISHU_APP_ID='你的飞书应用ID'
export FEISHU_APP_SECRET='你的飞书应用Secret'
export FEISHU_APP_TOKEN='飞书多维表格AppToken'
export FEISHU_TABLE_ID='飞书多维表格TableID'
⚠️ 不要硬编码凭证。run_pipeline.py 从环境变量读取。
config.json
定义账号类型、数据源、打分权重、推送时间。首次使用修改 config.json 中的 _env_required 对应值。
工作流
Step 1：抓取热点
使用浏览器访问 https://tophub.today/，抓取跨平台聚合热点。
重点关注：知乎热榜、微博热搜、36氪/虎嗅/少数派（科技/AI）、微信热文。
或一键运行：
# 设置环境变量后
python3 run_pipeline.py
Step 2：筛选与打分
5 维度打分（总分100）：
- 热度（30分）：跨平台聚合值
- 相关度（25分）：匹配账号定位（AI/效率/自媒体/副业）
- 时效性（20分）：事件新鲜度
- 可写性（15分）：素材丰富度、切入角度
- 差异化（10分）：竞品覆盖度、独特视角
筛选规则：≥70分推送 | 60-69分待评估 | \x3C60分不推送
Step 3：生成初步三要素
对每个高分选题生成粗粒度（~60%）三要素：
- 目标人群：谁会关注
- 痛点：戳中了什么
- 解决方案：从什么角度写
Step 4：写入飞书选题库
通过飞书 API 批量写入多维表格（run_pipeline.py 内置此逻辑）。
字段：选题标题、平台来源、热度值、采集时间、状态（待处理）、目标人群、痛点、解决方案、标签、优先级、备注。
Step 5：推送 Telegram
使用 message 工具推送选题卡片：
📊 选题推荐 | 2026-03-09
🔥 S级（强烈推荐）
1. [标题] — 热度XXX万 | 来源：知乎+微博
✅ A级（推荐）
2. [标题] — 热度XXX万
回复「采纳1 3 5」→ 笔尖专家开写
状态管理
- 创建记录 → "待处理"
- 用户采纳 → "已采纳"
- 用户放弃 → "已放弃"
触发方式
- 定时：每日 8:00 / 12:00 / 18:00
- 手动："帮我选题" / "获取热点" / "启动选题专家"
注意事项
- 素材必须严谨，不编造
- 每次最多推送 8 个选题
- 优先推荐与账号定位强相关的话题
- 记录用户的采纳/放弃偏好，持续优化推荐
变更日志
- v1.1.0 (2026-03-10)：凭证从硬编码改为环境变量；删除多余文档文件；config.json 脱敏。
- v1.0.0 (2026-03-10)：首次发布。tophub.today 热榜抓取、5维度打分、飞书选题库写入、Telegram 推送。
Usage Guidance
This skill's code matches its description (crawl热点, score, write to Feishu, push Telegram) but the package metadata omitted required environment variables and dependency info. Before installing: (1) review and be comfortable providing FEISHU_APP_ID/FEISHU_APP_SECRET/FEISHU_APP_TOKEN/TABLE_ID — create a dedicated Feishu app/account with minimal scopes and do not reuse high‑privilege tenant secrets; (2) confirm the Telegram push configuration (chat id/token) if you plan to enable pushes; (3) inspect the code (already included) for any endpoints you don't recognize and ensure network access to the listed sites is acceptable in your environment; (4) run it in an isolated/sandbox environment first and install required Python packages (requests, httpx, beautifulsoup4, feedparser, etc.); (5) be aware it will create files under ~/.openclaw/... and /tmp and will call external web services — if those behaviors are acceptable you can proceed, otherwise request the publisher to correct registry metadata and add a clear requirements.txt and explicit permission/scopes documentation.
Capability Analysis
Type: OpenClaw Skill
Name: inkroam-topic-expert
Version: 1.1.0
The skill bundle implements a comprehensive 'Topic Expert' workflow designed to scrape hot topics from various platforms (Weibo, Zhihu, 36kr, Tophub, etc.), score them based on relevance to AI/tech, and manage them via Feishu (Lark) and Telegram. The code is well-structured, uses standard libraries like BeautifulSoup and httpx for scraping, and follows security best practices by retrieving API credentials from environment variables rather than hardcoding them. No evidence of data exfiltration, malicious persistence, or prompt injection was found; all network and file operations are strictly aligned with the stated purpose of content discovery and management.
Capability Assessment
Purpose & Capability
The code implements exactly what the description says: crawling热点 sources, scoring, generating three elements, writing records to a Feishu bitable and pushing Telegram messages. However the registry metadata lists no required environment variables or binaries while SKILL.md and config.json (and run_pipeline.py) clearly require Feishu credentials (FEISHU_APP_ID, FEISHU_APP_SECRET, FEISHU_APP_TOKEN, FEISHU_TABLE_ID) and optionally TELEGRAM_CHAT_ID — the metadata omission is an inconsistency that should be corrected.
Instruction Scope
The runtime instructions and scripts keep to the stated scope: they fetch public web APIs/pages/RSS, perform local scoring/filtering, store items in a local SQLite DB, write to Feishu via its API, and push formatted messages to Telegram. The scripts read/write /tmp and create a DB under the user's home (~/.openclaw/...), which is expected for this kind of tool. There is no unrelated file-system or credential access requested in the instructions beyond the Feishu/Telegram creds.
Install Mechanism
There is no install spec (instruction-only at registry level) but the bundle contains many Python scripts that import third‑party libraries (requests, httpx, bs4/BeautifulSoup, feedparser). No dependency list / requirements.txt / install instructions are provided in the registry metadata or SKILL.md, so the runtime environment may be missing required packages; this is an operational gap and increases friction/risk if users attempt to run it without vetting the code and dependencies.
Credentials
The code legitimately needs Feishu API credentials to write to the bitable and optionally a Telegram chat ID/token to push messages; those are proportional to the described functionality. The concern is that the registry metadata declared no required environment variables/primary credential while the code will fail or prompt for sensitive credentials at runtime. Before providing credentials, users should verify Feishu app scopes and that credentials go to a dedicated app/account. The skill will use those credentials to obtain tenant tokens and call Feishu APIs (i.e., it can write data to a remote service).
Persistence & Privilege
The skill does not request 'always: true' and does not modify other skills or global agent settings. It can be run as a daemon and writes a local SQLite DB and log/PID files under the user's workspace/home; that persistence is appropriate for its function (stateful topic tracking).
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install inkroam-topic-expert - After installation, invoke the skill by name or use
/inkroam-topic-expert - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.1.0
Initial release. Real-time trending topic monitoring from 20+ platforms via tophub.today. 5-dimension intelligent scoring (popularity/relevance/timeliness/writability/differentiation). Auto-generates preliminary three-elements, writes to Feishu topic database and pushes to Telegram. Supports scheduled and manual trigger modes.
Metadata
Frequently Asked Questions
What is Topic Selection Expert?
选题专家：全网热点监控 → 筛选打分 → 初步三要素 → 写入飞书选题库 → 推送Telegram。Use when user asks to 启动选题专家 / 获取选题推荐 / 查看热点 / 帮我选题。 It is an AI Agent Skill for Claude Code / OpenClaw, with 342 downloads so far.
How do I install Topic Selection Expert?
Run "/install inkroam-topic-expert" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Topic Selection Expert free?
Yes, Topic Selection Expert is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Topic Selection Expert support?
Topic Selection Expert is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Topic Selection Expert?
It is built and maintained by inkroam (@bytearch1990-beep); the current version is v1.1.0.
More Skills