obsidian-content-importer - SKILL.md Agent Skill

name: obsidian-content-importer description: Use when a user gives one or many WeChat or Xiaohongshu links and wants the parsed notes saved to a specific local path without relying on the Obsidian plugin UI. This skill wraps the repository's import capability as a direct agent tool through the bundled script, including batch import, frontmatter generation, media download, and JSON result reporting.

Obsidian Content Importer

This skill is for direct execution, not for extending the plugin.

Humans can keep using the Obsidian plugin UI.
Agents should use this skill's script to import links directly.
The target path can be any local folder. If it is inside an Obsidian vault, the notes become available there automatically.

What this skill does

Given one link or a batch of links, the script will:

Detect supported platforms: WeChat or Xiaohongshu
Fetch and parse the page
Convert content into Markdown
Write note files into the requested output directory
Optionally download media into media/<title>/
Return a JSON summary with created files and failures

Use this script

Bundled executable: scripts/import-content.mjs

Run it with:

node scripts/import-content.mjs --output-dir "/absolute/output/path" --category "研究" --url "https://mp.weixin.qq.com/s/xxxx"

Batch import:

node scripts/import-content.mjs --output-dir "/absolute/output/path" --category "研究" --download-media --input-file "/tmp/links.txt"

Multiple URLs directly:

node scripts/import-content.mjs \
  --output-dir "/absolute/output/path" \
  --url "https://mp.weixin.qq.com/s/xxxx" \
  --url "https://www.xiaohongshu.com/discovery/item/xxxx"

Install check:

node ~/.codex/skills/obsidian-content-importer/scripts/import-content.mjs --help
node ~/.claude/skills/obsidian-content-importer/scripts/import-content.mjs --help

Agent workflow

Decide the output directory with the user or infer it from context.
Prefer repeated --url for a small batch; use --input-file for larger batches.
Add --download-media only when local media is actually needed.
Parse the JSON result and report:
- created note paths
- failed links and reasons
- skipped invalid lines

If the skill seems missing after installation:

Confirm the final path ends with obsidian-content-importer/SKILL.md rather than an extra nested directory.
Confirm scripts/import-content.mjs exists beside SKILL.md.
Start a fresh agent session and invoke the skill by name in the task.

Output contract

Notes: <output-dir>/<title>.md
Media: <output-dir>/media/<title>/...
Common frontmatter: platform, title, source, category, imported_at, cover, type
WeChat also writes: account, wechat_id, alias, author, published_at, published_ts, description