url-to-obsidian

name	url-to-obsidian
description	Capture one or more URLs into Obsidian notes. Use this when the user wants to save a URL, paste text containing URLs, or turn WeChat/GitHub/X/general web pages into Obsidian notes.

URL to Obsidian

Capture one or more URLs from the user's real Chrome session and save them as Obsidian notes.

First Use

Before running the skill, create url_to_obsidian.env from url_to_obsidian.env.example and set:

• URL_TO_OBSIDIAN_TARGET_DIR to an absolute Obsidian path

Workflow

The agent should treat this skill as export -> current session writes overrides -> generate. Do not rely on the scripts to make a second LLM call on their own.

1. Run export-only capture:
   • python3 scripts/url_to_obsidian.py export "<user input>"
2. Read the exported JSON from:
   • URL_TO_OBSIDIAN_SOURCE_JSON
   • default: ~/.dev-browser/tmp/url-to-obsidian-export.json
3. In the current agent session, create higher-signal title, summary, and tags for the exported items.
   • Keep one output entry per item key.
   • Reuse existing overrides when the file already exists.
   • Only create or update entries for URLs present in the current export JSON.
4. Write overrides to:
   • URL_TO_OBSIDIAN_LLM_OVERRIDES_FILE
   • default: ~/.dev-browser/tmp/url-to-obsidian-llm-overrides.json
5. Follow the JSON shape from:
   • url_to_obsidian_llm_overrides.example.json
6. Use the prompt template when writing overrides in the current session:
   • overrides_prompt_template.md
7. Generate final notes:
   • python3 scripts/url_to_obsidian.py generate

Agent Rules

• When the user asks to capture URLs, prefer the full agent-assisted flow instead of calling python3 scripts/url_to_obsidian.py in default full mode without overrides.
• The current agent session is responsible for writing the overrides JSON.
• If no overrides are needed, say that explicitly and then run the final generate step.
• Keep summaries concrete and compact; do not add filler.

Input Handling

• Accepts a direct URL
• Accepts pasted text containing one or more URLs
• Accepts piped text from stdin
• Extracts unique http:// and https:// URLs in appearance order

Supported Sources

• WeChat articles
• GitHub repositories, issues, pull requests, discussions, and README pages
• X / Twitter posts
• General web pages available in the user's current Chrome session

Output

• One note per captured URL under URL_TO_OBSIDIAN_TARGET_DIR
• 000 - URL 采集索引.md
• .url_to_obsidian_state.json