Ai News Skills

AI 新闻全自动工作流：采集公开网页新闻、生成日报与早报导读、合成周报趋势与深度解读。包含三个协作 skill：ai_news_fetcher（采集）、ai_news_reporter（日报）、ai_weekly_reporter（周报）。

Security Scan

The skill's code, instructions, and dependencies are coherent with its stated purpose (web-news fetch → Feishu bitable → daily/weekly reports); no disproportionate credentials or hidden endpoints were found, but it does send target URLs to the third‑party r.jina.ai reader and installs agent-reach via pip so review those behaviors before installing.

Purpose & Capability

Name/description (fetcher→reporter→weekly reporter) align with the included files: fetcher scripts parse web pages and produce Feishu bitable JSON, reporter builds daily reports from signals. The skill requires Feishu app tokens/table IDs (kept in local JSON examples) and Agent Reach/jina.ai for page-to-markdown conversion — these match the stated purpose.

Instruction Scope

Runtime instructions and scripts focus on fetching public webpages (via curl to r.jina.ai), normalizing/extracting fields, building bitable records, and generating reports; they read and write files under ~/.openclaw/workspace/skills and update local docs like watch_chains.md/analysis_framework.md. Important: URLs of target pages are sent to r.jina.ai (an external service) for markdown extraction — this is expected for public pages but is an external network disclosure to be aware of.

Install Mechanism

No opaque remote archive downloads. install.sh copies skill files into ~/.openclaw/workspace/skills and runs 'pip install agent-reach' and 'agent-reach install --env=auto'. Installing a PyPI package and running its installer is a moderate, expected action for this skill, but it modifies the user's Python environment and runs the agent-reach installer.

Credentials

The registry metadata lists no required env vars, and the skill uses local JSON config files (bitable_target.json, watch_target.json, wiki_target.json) to hold Feishu credentials (app_token, table_id, etc.). Requesting Feishu credentials is proportionate to writing to Feishu. Note: credentials are file-based (not declared as env vars in registry) — ensure these files are created locally and not shared with third parties or agent chat history.

Persistence & Privilege

always:false and no unusual privileged flags. Agents/openai.yaml sets allow_implicit_invocation:true (normal for skills). The skill will ask to create cron jobs and will update local watch/analysis files, but this is consistent with its function; there is no evidence it requests system‑wide configuration beyond the user's workspace.

Assessment

What to check before installing: - Confirm you are comfortable with target article URLs being sent to r.jina.ai (used to convert pages to markdown). Do not feed private/internal URLs. - The skill expects Feishu app credentials (app_token, table_id, etc.) stored in local JSON files under ~/.openclaw/workspace/skills; create a dedicated Feishu app with minimal scope and store credentials locally (do not paste them into public chat). - install.sh will run 'pip install agent-reach' and 'agent-reach install' which modifies your Python environment — review agent-reach's installer if you want to avoid global changes (consider a virtualenv). - The skill will read/write files in your OpenClaw workspace (including watch_chains.md and analysis_framework.md) and may create cron tasks via the agent — if you want tighter control, perform installation manually and inspect scripts first. - If you want extra caution: run the install and the first fetch in a sandbox or test workspace, inspect the (omitted/truncated) files that perform Feishu API interactions to verify they only use the local JSON credentials and to confirm the exact API endpoints and scopes used. If any of these points makes you uncomfortable (sending URLs externally, pip installs, automatic cron creation, or file writes), delay installation or audit the relevant scripts before proceeding.

一套完整的 AI 新闻自动化工作流，从采集到日报到周报。

公开网页 → 自动采集(fetcher) → 飞书原始新闻表 → AI筛选(reporter) → 日报 & 早报导读
                                                                      ↓ (每周日)
                                                          周报趋势 & 深度解读(weekly_reporter)
                                                                      ↓
                                                          更新产业链观察图谱

包含的 skill

Skill	职责	频率
skills/ai_news_fetcher	从公开网页采集 AI 新闻，写入飞书多维表格	每天 3-4 次
skills/ai_news_reporter	从飞书表筛选高价值事件，生成日报和早报导读	每天 1 次
skills/ai_weekly_reporter	基于本周日报合成趋势线，深度解读，更新观察链	每周日 1 次

使用时请指定具体 skill，例如 使用 ai_news_fetcher skill 采集新闻 或 使用 ai_news_reporter skill 生成日报。

前置依赖

• 飞书多维表格：需要配置 app_token 和 table_id，填入各 skill 的 bitable_target.json（参考 .example.json）
• Agent Reach：Agent Reach 提供网页读取能力
• Python 3.10+：运行数据处理脚本

凭证说明：本 skill 不内置任何飞书凭证或 API 密钥。所有配置文件（bitable_target.json、watch_target.json）由用户在本地创建，已被 .gitignore 排除。仓库中的 .example.json 仅包含占位符（YOUR_APP_TOKEN 等）。

第三方服务：使用 Jina Reader（免费、无需 API key）将公开网页转为 markdown 文本。采集和阅读原文时，目标网页的 URL 会发送到 r.jina.ai，不会发送飞书凭证或用户数据。

详细说明

每个 skill 的 SKILL.md 包含完整的执行流程、硬约束和模板，请参阅各自目录。