s-kostyaev/.emacs.d/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

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

1. Save the page as markdown (cleaned).

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

1. Inspect structure before reading content.

mq /tmp/result.md '.tree'

1. Extract only the needed sections.

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

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

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

1. 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.