How do I install the Web Summary skill?

Run npx skillmds@latest add dreamlab-ai/web-summary in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Web Summary safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Web Summary?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Web Summary free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Web Summary?

dreamlab-ai (@dreamlab-ai) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Web Summary by dreamlab-ai — Agent Skill

Name: Web Summary
Author: dreamlab-ai

Web Summary

Summarise a single web page or YouTube video into short/medium/long notes and extract semantic topic links for an Obsidian vault. Trigger when the user says "summarise this URL/article/video", "get the YouTube transcript", "pull the key points from this page", or "make vault/Obsidian topic links from this". Runs the `web-summary` subcommand of `agentbox-mcp` (a Rust rmcp server) that scrapes the URL and summarises via the Ontology Loom facade (model-swappable LLM door). NOT for interactive browser automation (use browser/playwright), NOT for multi-URL comparison or structured extraction (use web-researcher scrape_page or the browser sidecar), NOT for broad multi-source cited web search (use perplexity-research/web-researcher), and NOT for text you already hold locally (summarise it directly).

dreamlab-ai 1a9a662 2 files · 7.6 KB Updated Sep 17, 2026

File contents

Web Summary Skill

Single-URL content summarisation and topic extraction via the web-summary subcommand of agentbox-mcp — a single Rust rmcp binary (services/agentbox-mcp) that also serves the imagemagick and gemini-url-context skills. It replaces the former Python FastMCP server one-for-one, including its own direct YouTube transcript fetch (watch page + timedtext, no youtube-transcript-api dependency). It fetches the page (or YouTube transcript) and summarises it through the Ontology Loom facade — the load-bearing, model-swappable LLM door (agentbox ADR-051). The former Z.AI service on port 9600 is retired; see references/architecture.md.

When to use

Summarise one web article, blog post, or documentation page
Extract and summarise a YouTube video transcript
Generate semantic topic links for note-taking in the vault (Obsidian)
Produce short, medium, or long summaries; extract key concepts from text

When not to use

Interactive browser automation (clicking, filling forms, live scraping) — use the browser or playwright skills.
Multi-URL comparison or structured data extraction from URLs — use web-researcher (scrape_page / search_and_scrape) or the browser sidecar for JS-rendered pages.
Broad, multi-source web search with citations — use perplexity-research or web-researcher.
Content you already have as local text — summarise it directly without this pipeline.

Tools

Tool	Description
`summarize_url`	Summarise content from any single URL (web or YouTube)
`youtube_transcript`	Extract full transcript from a YouTube video
`generate_topics`	Generate semantic topic links from text
`health_check`	Verify Ontology Loom facade connectivity

Examples

# Summarise a web article
summarize_url({
    "url": "https://example.com/article",
    "length": "medium",
    "include_topics": True,
    "format": "obsidian"
})

# Get a YouTube transcript
youtube_transcript({"video_id": "dQw4w9WgXcQ", "language": "en"})

# Generate topic links
generate_topics({"text": "Your text content here...", "max_topics": 10, "format": "obsidian"})

Output formats

obsidian is the default (ADR-2028 D4) and emits - [[Topic]] wiki-links ready to paste into a vault page; plain emits - Topic.

logseq is legacy: it is still selectable and emits the identical - [[Topic]] wiki-link syntax, but it is retained only for callers pinned to the old name and should not be used in new work. Pages this skill's output lands in carry V2 YAML frontmatter — see project/docs/VAULT-corpus-format.md §V2.

Deeper reference

Architecture diagram, LLM backend wiring, environment variables (LLM_URL, LLM_MODEL, LLM_TIMEOUT), troubleshooting, and VisionClaw integration: references/architecture.md.

---name: web-summarydescription: Summarise a single web page or YouTube video into short/medium/long notes and extract semantic topic links for an Obsidian vault. Trigger when the user says "summarise this URL/article/video", "get the YouTube transcript", "pull the key points from this page", or "make vault/Obsidian topic links from this". Runs the `web-summary` subcommand of `agentbox-mcp` (a Rust rmcp server) that scrapes the URL and summarises via the Ontology Loom facade (model-swappable LLM door). NOT for interactive browser automation (use browser/playwright), NOT for multi-URL comparison or structured extraction (use web-researcher scrape_page or the browser sidecar), NOT for broad multi-source cited web search (use perplexity-research/web-researcher), and NOT for text you already hold locally (summarise it directly).---# Web Summary SkillSingle-URL content summarisation and topic extraction via the `web-summary`subcommand of `agentbox-mcp` — a single Rust `rmcp` binary(`services/agentbox-mcp`) that also serves the `imagemagick` and`gemini-url-context` skills. It replaces the former Python FastMCP serverone-for-one, including its own direct YouTube transcript fetch (watch page +timedtext, no `youtube-transcript-api` dependency). It fetches the page (orYouTube transcript) and summarises it through the Ontology Loom facade — theload-bearing, model-swappable LLM door (agentbox ADR-051). The former Z.AIservice on port 9600 is retired; see `references/architecture.md`.## When to use- Summarise one web article, blog post, or documentation page- Extract and summarise a YouTube video transcript- Generate semantic topic links for note-taking in the vault (Obsidian)- Produce short, medium, or long summaries; extract key concepts from text## When not to use- Interactive browser automation (clicking, filling forms, live scraping) — use the `browser` or `playwright` skills.- Multi-URL comparison or structured data extraction from URLs — use `web-researcher` (`scrape_page` / `search_and_scrape`) or the `browser` sidecar for JS-rendered pages.- Broad, multi-source web search with citations — use `perplexity-research` or `web-researcher`.- Content you already have as local text — summarise it directly without this pipeline.## Tools| Tool | Description ||------|-------------|| `summarize_url` | Summarise content from any single URL (web or YouTube) || `youtube_transcript` | Extract full transcript from a YouTube video || `generate_topics` | Generate semantic topic links from text || `health_check` | Verify Ontology Loom facade connectivity |## Examples```python# Summarise a web articlesummarize_url({ "url": "https://example.com/article", "length": "medium", "include_topics": True, "format": "obsidian"})# Get a YouTube transcriptyoutube_transcript({"video_id": "dQw4w9WgXcQ", "language": "en"})# Generate topic linksgenerate_topics({"text": "Your text content here...", "max_topics": 10, "format": "obsidian"})```## Output formats`obsidian` is the **default** (ADR-2028 D4) and emits `- [[Topic]]` wiki-linksready to paste into a vault page; `plain` emits `- Topic`.`logseq` is **legacy**: it is still selectable and emits the identical`- [[Topic]]` wiki-link syntax, but it is retained only for callers pinned tothe old name and should not be used in new work. Pages this skill's outputlands in carry V2 YAML frontmatter — see `project/docs/VAULT-corpus-format.md`§V2.## Deeper referenceArchitecture diagram, LLM backend wiring, environment variables (`LLM_URL`,`LLM_MODEL`, `LLM_TIMEOUT`), troubleshooting, and VisionClaw integration:[`references/architecture.md`](references/architecture.md).

Frequently asked questions