← 返回 Skills 市场
iret77

Non-Annoying News

作者 iret77 · GitHub ↗ · v0.2.1 · MIT-0
cross-platform ✓ 安全检测通过
91
总下载
0
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install non-annoying-news
功能描述
Create, configure, and render a source-grounded personal newspaper/news digest from URLs, bookmarks, X/Twitter bookmarks, browser reading lists/bookmarks, re...
使用说明 (SKILL.md)

Non-Annoying News

Create a personal newspaper that explains what matters without clickbait, dashboard cards, vague link summaries, or filler boxes.

Package version: v0.2.1.

Use modes

  • First-run setup: user wants the skill installed/configured, recurring issues, source adapters, X/bookmark inputs, naming, cadence, cron, delivery, or design choices.
  • One-off issue: user supplies URLs, notes, files, feeds, or a topic and asks for a digest now.
  • Recurring production: user already has a completed local config/signals and wants the next issue rendered and delivered.
  • Demo issue: user explicitly asks for a generic demo/sample, not a personalized newspaper.

Required workflow

  1. Load references/editorial-standard.md before selecting stories or writing an issue.
  2. For first-run setup, missing/incomplete config, recurring setup, cron/schedule, or customization requests, load references/onboarding.md, references/config-schema.md, references/source-adapters.md, and references/design-presets.md.
  3. Apply the Personalization Gate from references/onboarding.md: do not create a personal issue, recurring config, delivery automation, or cron job until the required onboarding choices are answered or explicitly confirmed by the user.
  4. Before rendering or delivering a PDF/HTML issue, load references/layout-and-render-qa.md; load references/design-presets.md if any visual choice is involved.
  5. Use elevated reasoning/Thinking when available (xhigh recommended) for onboarding synthesis, story selection, editorial proof-reading, and final QA.
  6. Gather user-provided sources first. Then use configured signal sources only when credentials/tools already exist or the user provides exported data.
  7. Treat bookmarks, reading-list entries, likes, saved items, and read-later queues as signals of intent, not automatically verified facts. Fetch/read linked content where possible and mark access limits.
  8. Write every article so it is understandable without seeing the original source, bookmark, or thread.
  9. Render HTML first, then PDF when possible. Create PNG previews/screenshots for every page and inspect them before delivery.
  10. If PDF rendering is unavailable, deliver the HTML and state the missing renderer clearly. Do not fake a PDF.

Personalization Gate

When a user asks to install, configure, “set this up”, create a recurring digest, or make their personal newspaper for the first time, the agent must start with onboarding questions. It may propose defaults, but it must not silently assume:

  • newspaper title/subtitle/language;
  • reader promise and intended use;
  • topics and exclusions;
  • source/signal mix, including whether to use X/Twitter bookmarks, browser reading lists/bookmarks, read-later apps, RSS/newsletters, web search, and pasted URLs;
  • cadence/frequency and whether a cron/scheduled job should be created;
  • delivery mode/channel and output format;
  • design preset, density, page count, and image policy.

If the user wants speed, offer a concise suggested config and ask for confirmation/corrections before saving it or generating the first personalized issue. Never create a cron job, external delivery, or credentialed-source workflow without explicit approval.

Exception: if the user explicitly asks for a generic demo/sample, you may generate a clearly labeled demo issue with sample content. Do not present it as personalized.

Non-negotiable output standard

  • The issue must read like a compact newspaper or magazine, not a link dump.
  • Each issue needs an editorial spine: why these stories belong together now.
  • Every article must contain: concrete event/claim, mechanism or argument, relevance/consequence, source/evidence boundary.
  • No internal process language inside the newspaper: no TODOs, queues, QA notes, placeholders, automation commentary, or excuses.
  • No filler boxes to patch layout holes. Empty side columns and lower thirds are layout failures; fix text or layout and render again.
  • Sources must be cited discreetly under articles or in a source note. Links are proof, not the content.

Recommended files and scripts

  • references/onboarding.md: first-run setup flow, required Personalization Gate, and user-facing questions.
  • references/config-schema.md: local configuration shape and copyable starter config.
  • references/source-adapters.md: portable source/signal options including X bookmarks and browser reading lists.
  • references/design-presets.md: newspaper naming, visual presets, and personalization knobs.
  • assets/config.example.json: starter config the agent may copy into a local project outside the public skill; it is intentionally incomplete until onboarding is done.
  • assets/newspaper-template/base.css: print CSS with theme/density tokens.
  • assets/newspaper-template/issue-template.html: minimal issue structure with balanced side rails.
  • scripts/check_config.py: validate local config before a recurring run.
  • scripts/collect_sources.py: normalize pasted URL/title/note lists into a source manifest.
  • scripts/render_issue.mjs: render HTML to PDF and PNG previews with Playwright when available.
  • scripts/qa_text.py: scan visible HTML/text for common editorial and privacy failures.

Privacy and portability

This skill is deliberately generic. Do not include user-specific names, accounts, private topics, local paths, workspace memories, handles, credentials, or channel IDs in reusable skill files or generated public examples. Put user-specific settings in a local project config outside the skill.

Delivery checklist

Before final answer:

  • Confirm onboarding is complete or explicitly state this is only a generic demo/sample.
  • Confirm PDF/HTML artifact path exists.
  • Confirm previews/screenshots were inspected page by page.
  • Confirm text QA and config validation passed when applicable.
  • Attach with an absolute MEDIA: path when the channel supports attachments.
  • Mention any source access limitations in the message, not as an excuse inside the issue.
  • If a publish/send/schedule action beyond delivering a local artifact is requested, require explicit approval.
安全使用建议
Before installing, decide which sources you actually want to use. Start with pasted URLs or exports if you are cautious. Do not paste secrets into chat, review any local config before saving it, approve Playwright/Chromium installation only if you need PDF rendering, and create a cron or external delivery workflow only after confirming the schedule, sources, and destination.
功能分析
Type: OpenClaw Skill Name: non-annoying-news Version: 0.2.1 The skill bundle is a well-structured tool for generating personalized newspapers from various digital signals. While it possesses high-risk capabilities such as task scheduling (cron) and processing sensitive data (browser bookmarks, X/Twitter bookmarks), it includes robust safety instructions and a dedicated privacy-focused script (`scripts/qa_text.py`) designed to detect and block the inclusion of local file paths or private identifiers in the output. The instructions in `SKILL.md` and `references/onboarding.md` explicitly mandate user consent for automation and warn against requesting credentials in chat, demonstrating a clear focus on privacy and user control rather than malicious intent.
能力评估
Purpose & Capability
The skill's personal newspaper purpose coherently matches its source collection, editorial QA, and rendering capabilities; the noteworthy part is that it can optionally use personal bookmarks, reading lists, newsletters, and delivery channels.
Instruction Scope
The instructions repeatedly require onboarding, user confirmation, and explicit approval before creating cron jobs, credentialed-source workflows, external delivery, or personalized recurring issues.
Install Mechanism
There is no automatic install spec, and the included helper scripts are visible and purpose-aligned; PDF rendering may require the user or agent to install Playwright/Chromium.
Credentials
No credentials or environment variables are required by default. Optional local profiles, APIs, mailbox tools, and delivery tools are proportionate to the stated purpose but should remain explicitly user-selected.
Persistence & Privilege
Recurring cron/scheduled operation is supported, but the artifacts clearly gate it on explicit confirmation of cadence, sources, delivery, and automation behavior.
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 non-annoying-news
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /non-annoying-news 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
v0.2.1
v0.2.1: tighten first-run onboarding with a mandatory Personalization Gate before personal issues, cron jobs, or external delivery; starter config now stays incomplete until confirmed.
v0.2.0
v0.2: guided onboarding, X/bookmark/reading-list signals, config validation, design presets, and stronger editorial/layout QA
元数据
Slug non-annoying-news
版本 0.2.1
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 2
常见问题

Non-Annoying News 是什么?

Create, configure, and render a source-grounded personal newspaper/news digest from URLs, bookmarks, X/Twitter bookmarks, browser reading lists/bookmarks, re... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 91 次。

如何安装 Non-Annoying News?

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

Non-Annoying News 是免费的吗?

是的,Non-Annoying News 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Non-Annoying News 支持哪些平台?

Non-Annoying News 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 Non-Annoying News?

由 iret77(@iret77)开发并维护,当前版本 v0.2.1。

💬 留言讨论