Token导航 LogoToken导航

Agent Skill · 研究检索

instagram-extract(Instagram 摘录)

instagram-extract 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。

instagram-extract 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为操作浏览器;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 65,351。

总安装 282周安装 12GitHub Stars 公开资料未说明下载 99核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

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:instagram-extract(Instagram 摘录)
来源仓库:https://github.com/michailbul/laniameda-skills
仓库路径:skills/instagram-extract
安装命令:
npx skills add https://github.com/michailbul/laniameda-skills --skill instagram-extract
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/michailbul/laniameda-skills --skill instagram-extract

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

instagram-extract 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词快速定位候选结果时使用。适用于内容研究、数据调研或信息聚合等需要从多个来源提取线索的任务场景。

通过 npx skills add 命令安装,需确认权限范围和是否触发联网或文件读写操作。建议结合原始 README 核验具体用法,并检查仓库维护状态与安全性。使用前应评估是否会执行命令或访问外部资源,避免意外行为。

来源仓库内路径:skills/instagram-extract

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/michailbul/laniameda-skills --skill instagram-…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要操作浏览器,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

instagram-extract 怎么安装?

在终端执行 npx skills add https://github.com/michailbul/laniameda-skills --skill instagram-extract。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

instagram-extract 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

instagram-extract 需要什么权限?

权限级别是操作浏览器。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。

instagram-extract 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

instagram-extract 用的人多吗?

公开数据是累计安装 282 次,近一周安装 12 次,下载 99 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

instagram-extract 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
instagram-extract

Instagram 摘录 · 本页

282—操作浏览器未明确有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Instagram Extract

Use Instagram as a “find + capture” stream: the user pastes a post link and says something generic like “save it”, and you decide what’s worth capturing, how to summarize it, and where it belongs.

The workflow should be agentic (use judgment) rather than deterministic.

Hard constraints:

  • Use the logged-in browser relay (browser tool with profile="chrome") when available.
  • Classify + save into the right pillar so it’s reusable later.
  • Report back in bullet points (no full transcript unless asked).
  • After saving, notify the media agent by appending a compact JSON record to the shared inbox file (details below).

Scope constraints (current reality):

  • This skill is for Instagram + Threads + LinkedIn (logged-in via Browser Relay / existing Chrome/Brave profile).
  • X/Twitter: don’t try to force it in Brave if it’s blocked; treat as out-of-scope for this skill (we’ll use a different approach/tool later).

Pillars (classification targets)

Pick the single best fit (canonical save) and optionally add a secondary tag when it overlaps. Do not duplicate saves across pillars unless the user explicitly asks.

Current pillars Michael wants to save:

Use these pillar ids in the media-agent inbox payload:

  1. Prompts — creating images (prompts_image)
  2. Tutorials/knowledge — marketing workflows (tutorials_marketing) — frameworks, copywriting, growth, positioning
  3. Design systems / design inspiration (design_inspo) — UI patterns, components, typography, layout
  4. Prompts — web design / development (prompts_webdev) — UI prompts, landing page prompts, component prompts

Also include optional tags like: storytelling, hooks, carousel, landing_page, ui, prompt_template, checklist, case_study.

Lightweight workflow (guidance, not a script)

1) Get access to the real post (relay)

  • Try to use browser relay takeover (profile="chrome").
  • If relay isn’t attached, ask the user to attach it (toolbar icon → badge ON).
  • If token is required, have the user set the gateway token in extension Options.

2) Extract enough to be useful

Prioritize signal over completeness.

For Instagram/Threads:

  • Caption: prefer og:description / og:title.
  • Carousel: grab slide text via image alt text (fast); OCR only when the slide text is garbled or crucial.

For LinkedIn:

  • Expand “see more” if present and extract the full post text.
  • If it’s a document carousel or images and text extraction is limited, still capture the media URLs + summarize what’s visible.

For all platforms:

  • Links: collect outbound URLs/domains if present.
  • Media: collect image URLs when available (good enough for downstream reuse).

3) Decide what matters (summarize)

Write 5–10 bullets focusing on:

  • the core idea / framework
  • anything immediately actionable
  • any reusable templates (hooks, structures, checklists)
  • any concrete examples worth saving

4) Decide what to do with it (agentic routing)

You’re optimizing for future reuse, not perfect capture.

Default behavior Michael wants:

  • If the user says "save it" and the post is weak/low-signal (generic motivation, no reusable artifact), ask once: “Not worth saving (too generic). Save anyway?”
  • If the user says "force save" / "save anyway": save regardless.
  • If the user says "light save": save a tiny note (title + link + 3 bullets) even if it’s generic.

Use these heuristics (not rules):

  • If it’s directly reusable instructions (a prompt, a template, a checklist, a swipeable structure) → save it.
  • If it’s vibes with no payload (generic motivation, no concrete steps) → don’t over-save; create a tiny note only if the framing is uniquely good.
  • If it’s ambiguous, ask *one* clarifying question max (e.g., “Do you want this saved as marketing knowledge or as design inspiration?”). Otherwise, pick your best guess and proceed.

Routing:

  • Prompt content (copy/paste generation instructions) → ingest via prompt-kb.
  • Tutorials/knowledge (marketing workflows) → local KB note.
  • Design systems / design inspiration → local KB note focused on patterns.
  • If it contains both a reusable prompt and a useful framework → save both (local KB summary + prompt-kb ingest).

5) Save (create durable artifacts)

A) Local KB save (tutorials/knowledge/design inspiration)

Save one canonical note in a stable, platform-agnostic structure:

  • kb/pillars/<pillar_id>/YYYY/MM/<YYYY-MM-DD>-<short-slug>.md

Where pillar_id is one of:

  • tutorials_marketing
  • design_inspo (If it’s a prompt-first asset, you can still save a short KB note, but the canonical storage may be prompt-kb.)

Also store a raw capture (optional but recommended when extraction is messy) in:

  • kb/inbox/<platform>/YYYY/MM/<YYYY-MM-DD>-<short-slug>.md

<platform> is instagram | threads | linkedin.

Note format (keep it tight):

  • Title (human readable)
  • Source URL
  • Platform (instagram|threads|linkedin)
  • Pillar (primary) + optional tags (secondary)
  • 5–10 bullet summary (your synthesis)
  • Why it matters (1–2 bullets)
  • Reusable templates/checklists (if any)
  • Links found (if any)
  • Media (image URLs if available)
  • Optional: extracted text only when it’s actually useful later

Slug guidance:

  • Prefer something that will be searchable later: <topic>-<creator> or <framework>-<topic>.
B) Prompt vault ingest (prompt-kb)
  • Use prompt-kb.
  • Tagging:

- Always include one: prompts | tutorials | resources | ideas - Add image_gen / video_gen / ui_design when appropriate - Add tool/model tags if known

  • Multiple slides/images for the same prompt template → shared promptIngestKey.
C) Media-agent handoff (always)

Append a single JSON line to: kb/_media_agent_inbox/ingest.jsonl

Schema (keep it compact):

{
  "createdAt": "YYYY-MM-DDTHH:mm:ssZ",
  "platform": "instagram|threads|linkedin",
  "url": "...",
  "pillar": "prompts_image|tutorials_marketing|design_inspo|prompts_webdev",
  "tags": ["optional", "tags"],
  "title": "...",
  "summary": ["...", "..."],
  "links": ["..."],
  "media": ["https://..."],
  "repurpose": {
    "x": {"hook": "...", "outline": ["...", "..."]},
    "carousel": {"slideTitles": ["...", "..."]},
    "shortVideo": {"beats": ["hook...", "beat...", "cta..."]},
    "cta": "..."
  },
  "saved": {
    "kbPaths": ["kb/..."],
    "promptKb": {"ingestKey": "optional", "tags": ["..."]}
  }
}

Implementation hint: use a tiny Python one-liner to append JSONL so you don’t need to read/overwrite the file.

6) Telegram report (bullet points only)

Return:

  • what it is (1 line)
  • 5–10 key takeaways
  • why it matters (1–2 bullets)
  • links found (if any)
  • what got saved + where (file path and/or “ingested to Convex”) + pillar + tags
  • Repurpose Pack (ALWAYS) — written in Michael’s voice (blunt, marketing-first, specific):

- X post: 1 hook + 3–5 bullet outline - Carousel: 6–8 slide titles - Short video: 15–30s script beats (hook → beats → CTA) - CTA: suggested comment keyword / next step

平台分布

Codex

36.75%
按下载量换算36

Claude

31.55%
按下载量换算31

Cursor

18.58%
按下载量换算18

Gemini CLI

10.05%
按下载量换算10

安全审计

Gen Agent Trust Hub

可疑

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills