# Web Browse Context

> Efficiently capture web pages as clean markdown and query them with `mq` to avoid loading full content into context. Use when browsing web pages, extracting specific sections, or summarizing sources with minimal context usage.

- Skill: `s-kostyaev/web-browse-context` (Agent Skill, multi-file: 4 files)
- Install (CLI): `npx skillmds@latest add s-kostyaev/web-browse-context`
- Raw SKILL.md: https://api.skillmd.com/api/skills/s-kostyaev/web-browse-context/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Docs & Writing
- Author: s-kostyaev (https://skillmd.com/u/s-kostyaev)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/s-kostyaev/web-browse-context

---


# Web Browsing With Context Economy

## Overview

Capture web pages into local markdown, inspect structure with `mq`, and extract only the relevant sections to keep context small and focused.

## Workflow

0. Use shell_command tool for every command from this skill.

1. Save the page as markdown (cleaned).

```bash
./scripts/url-to-md 'https://example.com' --clean-content > /tmp/result.md
```

2. Inspect structure before reading content.

```bash
mq /tmp/result.md '.tree'
```

3. Extract only the needed sections.

```bash
mq /tmp/result.md '.section("Using Blueprints") | .text'
```

4. Search when you are not sure where the content lives.

```bash
mq /tmp/result.md '.search("auth")'
```

5. If the search returns a large subtree, narrow it by re-running `.tree` on the specific section(s), then extract `.text`.

## Query Patterns

- View a fuller outline: `mq /tmp/result.md '.tree("preview")'`
- Get all headings: `mq /tmp/result.md '.headings'`
- Extract code blocks by language: `mq /tmp/result.md '.code("bash")'`
- List links: `mq /tmp/result.md '.links'`
- Extract a section: `mq /tmp/result.md '.section("Name") | .text'`

## Context Hygiene

- Never `cat` the whole markdown file into the conversation.
- Prefer `.tree` and `.search` to locate content, then extract only the minimal `.text` needed.
- Use one file per page (e.g., `result-pricing.md`, `result-api.md`) to keep sources separated.
- If a page changes, re-run the capture command and overwrite the corresponding file.
- Use `mq --help` whenever you need more selectors or flags.