Token导航 LogoToken导航TokenDH.com
研究检索操作浏览器github未标认证来源可访问许可证需确认审计提醒

hn-to-x-posterhn 到 x 海报

Agent Skill

hn-to-x-poster 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

259

周安装

11

GitHub Stars

公开资料未说明

下载量

91
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:hn-to-x-poster(hn 到 x 海报)
来源仓库:https://github.com/hexbee/hello-skills
仓库路径:skills/hn-to-x-poster
安装命令:
npx skills add https://github.com/hexbee/hello-skills --skill hn-to-x-poster
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/hexbee/hello-skills --skill hn-to-x-poster

简介

用于查找、检索和筛选相关信息。

  • 适合根据关键词、任务场景或来源线索快速定位候选结果。
  • 通过 npx skills add 命令从指定仓库安装并使用。
  • 建议确认权限范围和维护状态,注意是否会触发联网或文件读写操作。
  • hn-to-x-poster 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

HN to X Poster

Overview

Use this skill for a browser-driven workflow that reads the current Hacker News front page, filters the top 30 stories for software or AI relevance, compresses the signal into a Chinese thread, and publishes it to x.com.

This skill is for direct execution, not just drafting. Default behavior is to post automatically if x.com is already logged in. If posting is blocked, fall back to a ready-to-send Chinese draft or thread and explain the blocker briefly.

When To Use

Use this skill when the user asks for any combination of:

  • Hacker News or HN top stories
  • HN top 30, front page, or current HN software or AI items
  • posting, tweeting, sending, or publishing the result to x.com or Twitter
  • a Chinese summary of current HN software or AI topics that should be posted automatically

Do not use this skill when:

  • the user only wants HN analysis without posting
  • the user wants deep article research instead of front-page signal extraction

Required Tooling

Use chrome-devtools MCP for the browser workflow.

Do not rely on API access to Hacker News or X unless the user explicitly changes the method.

Workflow

1. Open The Two Sites

  • Open or switch to https://news.ycombinator.com/
  • Open or switch to https://x.com/
  • Confirm that X is logged in before attempting to post

If X is not logged in, stop automatic posting and return a Chinese draft the user can publish manually.

2. Read HN Top 30

  • Read the current HN front page
  • Capture the first 30 ranked stories only
  • Use the front-page title and visible context as the default source of truth

Do not open every article by default. This skill is meant to summarize the HN front-page signal, not to perform full article research.

3. Filter For Software Or AI Relevance

Keep items that are clearly about:

  • software engineering
  • programming languages
  • developer tools
  • operating systems
  • networking, infrastructure, databases, compilers, security, or systems work
  • AI, LLMs, agents, model tooling, MCP, inference workflows, or AI product/platform shifts

Usually drop items that are mainly:

  • general politics
  • culture or biography
  • pure science without a software or AI angle
  • consumer oddities with no developer or AI relevance

If an item is borderline, prefer keeping it only when a technical audience would plausibly care.

4. Plan Post Structure

Decide the thread structure before writing.

If the user explicitly specifies the number of posts in the thread:

  • follow the user's specified count
  • organize the content to fit that count as cleanly as possible
  • do not override the count just because a different number seems better

If the user does not specify the number of posts:

  • default to a thread of at least 3 posts
  • treat HN front-page signal as high-density by default; do not assume one post is enough
  • use a total -> points -> takeaway structure as the baseline shape
  • allow 4+ posts whenever that makes the thread clearer
  • do not optimize for the fewest posts; optimize for explaining the signal cleanly
  • split by natural content blocks instead of arbitrary equal lengths
  • prefer adding another post over over-compressing one crowded post

Default thread planning heuristics:

  • 3 posts: the default minimum when the user did not specify a count
  • 4 posts: use when signals naturally separate into multiple clusters, such as AI, security/systems, languages/tools, and conclusion
  • 5+ posts: acceptable when needed to keep each post readable and specific
  • 1 or 2 posts: only use when the user explicitly asks for that count

When writing a thread, give each post a distinct job. Typical default structure:

  • post 1: overall read on today's HN board, set the tone
  • post 2: first major cluster of signals
  • post 3: second major cluster plus closing takeaway

If using more than 3 posts, keep the same principle:

  • opening post sets the frame
  • middle posts each cover one clear cluster
  • final post lands the synthesis, verdict, or forward-looking take

5. Build The Post

Write the final Chinese content as a thread that:

  • sounds natural on X
  • highlights only the strongest 4 to 8 software or AI signals
  • compresses repeated themes instead of listing everything
  • ends with a clear conclusion about the pattern

Default style:

  • concise
  • readable by Chinese tech audiences
  • opinionated and human, not robotic
  • may use slang, idioms, colloquial turns of phrase, or other distinctive personal expressions when they improve voice
  • keep the voice vivid but still grounded in what was actually visible on HN
  • no hashtags unless the user asks
  • no links unless the user asks

If writing a thread:

  • keep each post readable on its own
  • maintain flow across posts instead of making each one feel like a separate unrelated update
  • use simple numbering such as 1/, 2/, 3/ when that improves clarity
  • avoid stuffing every retained HN item into the thread; keep only the strongest signals
  • prefer sharp summaries over flat inventory lists
  • let the wording have some personality, but do not let jokes or slang blur the actual point

6. Enforce Length Before Posting

Before posting, make sure each post is short enough for a standard X post.

Length rules to enforce:

  • standard X posts are limited to 280 characters for normal users
  • a Chinese character usually consumes about 1-2 characters of the counter
  • a link usually consumes about 23 characters even when it looks shorter
  • always trust the visible X counter over any rough manual estimate

If it is too long:

  • first consider splitting the content into more thread posts
  • compress only after the thread structure is already reasonable
  • remove low-signal items first
  • collapse repeated categories into one phrase
  • keep the main pattern, not the full inventory

For threads:

  • shorten each post independently
  • keep the thread balanced, but do not force equal lengths
  • if one post is doing too much, split it before flattening it
  • if the thread is getting too long overall, remove lower-signal details before sacrificing the total-summary-total structure

7. Post On X

Use the homepage composer with keyboard-first actions.

For a single post:

  • go to https://x.com/home
  • prefer pressing n first to open the new post composer
  • if n does not open the composer, fall back to clicking the homepage Post button
  • confirm the composer is visible before typing
  • NEVER use fill - must use type_text to input the complete text in one go
  • Do not input in steps or modify mid-way - input the full Chinese text at once
  • verify both of these before posting:

- the post button is enabled - the character counter looks normal

  • prefer pressing Ctrl + Enter to publish
  • if Ctrl + Enter does not publish, fall back to clicking Post

For a thread:

  • go to https://x.com/home
  • prefer pressing n first to open the new post composer
  • if n does not open the composer, fall back to clicking the homepage Post button
  • confirm the composer is visible before typing
  • NEVER use fill - every post in the thread must be entered with type_text
  • type the full text of post 1 in one go
  • find and click the Add post button to create the next composer cell
  • type the full text of each later post in one go with type_text
  • keep repeating until all planned posts are present
  • verify the final publish control is enabled and the visible counters look normal
  • prefer pressing Ctrl + Enter once to publish the entire thread
  • if Ctrl + Enter does not publish, fall back to clicking the final Post all or equivalent publish button

8. Verify Completion

After posting:

  • go to the account profile or resulting post view
  • for a single post, confirm the new post appears as the latest visible post
  • for a thread, confirm the newly published posts appear as the latest visible posts in the expected order
  • capture the first post URL if available
  • when possible, verify at least one reply-linked or sequential thread post is also visible, not just the first post

Do not claim success unless the post or thread is visibly present in the browser.

Failure Handling

X Not Logged In

Do not attempt login on the user's behalf unless they explicitly ask.

Return:

  • a brief note that automatic posting is blocked because X is not logged in
  • the final Chinese draft or thread draft

Composer Does Not Recognize Input

If text appears in the textbox but the post button stays disabled:

  • NEVER use fill - must use type_text
  • treat this as an input-recognition failure, not a writing failure
  • Press Ctrl+A to select all, then use type_text to input the complete text again in one go
  • re-check both the post button and the character counter before posting

Composer Counter Looks Wrong

If the text length is obviously short enough but X claims it is over limit or otherwise shows a broken counter:

  • assume the composer state is corrupted
  • refresh or reopen https://x.com/home
  • press n again to reopen a fresh composer when possible
  • re-enter the final text in the homepage modal composer
  • do not keep retrying in the broken composer state

Post Or Thread Is Too Long

Restructure, shorten, and retry. Do not stop at the first over-limit failure.

If using a thread:

  • first split overloaded posts into additional thread posts when that improves readability
  • first shorten the affected post
  • then remove lower-signal details
  • increasing thread length is acceptable when it makes the post sequence clearer and the user did not specify the count

Posting Still Fails

Return:

  • the final Chinese text or thread
  • the exact blocker, such as disabled post button or missing login state

Output Expectations

When successful, give a short result that includes:

  • that HN top 30 was reviewed
  • that software or AI items were selected
  • whether the result was a single post or a thread
  • that the Chinese post or thread was published to X
  • the first X post URL

When blocked, give:

  • a brief reason
  • the final Chinese draft or thread draft

Operating Notes

  • Treat HN as a front-page signal source, not a complete market map
  • Prefer a small set of strong signals over exhaustive coverage
  • Favor execution reliability over perfect prose
  • If the user asks for a draft only, do not auto-post
  • On X, use the homepage modal composer as the standard posting path, with n as the preferred way to open it
  • Do not trust visible text alone; the post button state and counter state are the real readiness checks
  • Prefer Ctrl + Enter as the standard publish action, with button click only as fallback
  • When the user specifies the number of thread posts up front, follow that instruction exactly
  • When the user does not specify a count, default to at least 3 posts and expand further when needed to explain the signal properly
  • HN usually contains enough density that a single-post summary is an exception, not the default
  • After publishing, verify on profile that the new post or thread is actually published before reporting success

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

35.7%
按下载量换算32

Claude

29.04%
按下载量换算26

Cursor

22.01%
按下载量换算20

Gemini CLI

11.14%
按下载量换算10

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills