跳到正文
cccskills
免费在 GitHub 公开

openmark-bookmark-dive

Take ONE specific bookmark URL and explore the neighborhood around it — tags, SIMILAR_TO neighbors, community peers — plus optionally WebFetch the page itself to summarize it. Use whenever Ahmad pastes a single URL and asks "what's near this", "tell me about this one", "expand this bookmark", "dig into this", "what else did I save like this", "give me the context for this URL". Triggers on "expand", "neighbors of", "what else like this", "dig into this URL", "context for this", and any message that's just a URL plus a short prompt.

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(1)

  • SKILL.md3.2 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

OpenMark — Bookmark Deep Dive

Anchor on ONE URL. Surface its graph neighborhood. Optionally summarize the page itself. Output a compact briefing.

When to use

User pastes a URL plus a short ask:

  • "what else did I save like [https://example.com/agents]"
  • "dig into [URL]"
  • "expand this bookmark: [URL]"
  • "give me the neighbors of [URL]"
  • "what's around this in OpenMark"

If the user gives no URL, fall back to openmark-fast-search. If the user gives a URL AND a broad topic, use openmark-deep-research.

Tool sequence

  1. Confirm the bookmark exists:

    • get_bookmark_full(url=<URL>)
    • If note == "Bookmark not found.", say so and stop. Don't pretend.
  2. Graph expand (this is the core call):

    • graph_expand(url=<URL>) — returns text describing tags, SIMILAR_TO list, community peers.
  3. WebFetch the URL itself (parallel to step 2 if possible):

    • WebFetch(url=<URL>, prompt="Summarize in 3-4 sentences. Then list 2-3 specific claims or numbers that are central to the page.")
    • If fetch fails (paywall, JS-only, 404), skip and note "Could not fetch live page."
  4. Optional broaden — only if user explicitly asks "what else":

    • For top 2 tags from get_bookmark_full's tags[], run find_by_tag(tag=<tag>, n=10).
    • For the community_id if present, run search_by_community(query=<title or first tag>, n=15).

Output format

# {Title}
{URL}

**OpenMark metadata**
- Source: {source}  ·  Category: {category}  ·  Score: {score}
- Tags: {tag1, tag2, tag3, ...}
- Created: {created_at if available, else "—"}

**What the page says** (from WebFetch)
{3-4 sentence summary + 2-3 specific claims}

**Similar bookmarks** (from SIMILAR_TO)
1. {title} — {url}
2. {title} — {url}
3. {title} — {url}
...

**Community peers** (from IN_COMMUNITY)
1. {title}
2. {title}
...

**Broader neighborhood** (only if requested)
{Top 10 from tag/community expansion}

Rules

  • The anchor URL appears at the top, exactly once, exactly as provided. No normalisation in the output (even if normalised internally for lookup).
  • Similar/peer URLs come from tool output, never invented. If the SIMILAR_TO list is empty, write "No SIMILAR_TO neighbors yet (run graph_hygiene if you just injected new items)."
  • WebFetch summary is the ONLY place free prose appears. Everything else is structured.
  • Don't editorialize on the page content. Just relay what WebFetch returned. Ahmad will form his own take.
  • If community_id is None on the node, say so — "Not yet in a Louvain community (GDS plugin not installed / not re-run)."

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Reusable skill for coding agents to build mediahost integrations, API clients, scrapers, bulk downloaders, and media tools. Incorporates the unofficial Python wrapper, auth flow, endpoints, rate-limit handling, and cross-language patterns ...

中文简介正在准备中。当前显示原文说明。

OthmanAdi/OpenMark102026年7月28日 更新

Reusable skill for coding agents (Claude, GPT, etc.) to build, optimize, debug, and productionize AI agents and coding agents in Rust. Embeds Rig/ADK patterns, Tokio async architecture, tracing observability, performance techniques, erro...

中文简介正在准备中。当前显示原文说明。

OthmanAdi/OpenMark102026年7月28日 更新

Remove AI-generated writing patterns from Egyptian Arabic (عامية مصرية / Masri) text. Use when editing or reviewing Egyptian dialect text to make it sound authentically human-written.

中文简介正在准备中。当前显示原文说明。

OthmanAdi/OpenMark102026年7月28日 更新

Remove AI-generated writing patterns from Modern Standard Arabic (MSA/الفصحى) text. Use when editing or reviewing Arabic formal text to make it sound naturally human-written.

中文简介正在准备中。当前显示原文说明。

OthmanAdi/OpenMark102026年7月28日 更新

Remove AI-generated writing patterns from Levantine Arabic (الشامي — Syrian, Lebanese, Palestinian) text. Use when editing or reviewing Levantine dialect text to make it sound authentically human-written.

中文简介正在准备中。当前显示原文说明。

OthmanAdi/OpenMark102026年7月28日 更新

Remove AI-generated writing patterns from Modern Hebrew (עברית מודרנית) text. Use when editing or reviewing Hebrew text to make it sound naturally human-written.

中文简介正在准备中。当前显示原文说明。

OthmanAdi/OpenMark102026年7月28日 更新

查看 OthmanAdi 的全部技能

报告此技能的问题