import

name	import
description	Import external resources (articles, PDFs, videos, podcasts) into the Obsidian vault using the summarize CLI. Generates a French summary, auto-tags, and stores in Resources/. Use this skill when the user wants to save, import, capture, or archive any external content (URL, PDF, video, podcast) into the vault.
argument_hint	<url_or_path> [#tags]

Import Resource

Import external content into the vault as a structured Obsidian note with French summary, auto-generated tags, and proper wikilinks.

Trigger

/import <url_or_path> [#tags] [--raw]

Input Parsing

1. Extract the URL or file path (first argument that looks like a URL or path)
2. Extract user-provided #tags (any #word tokens)
3. Detect --raw flag: if present, store the extracted raw content instead of a summary

Workflow

Step 1 — Extract raw content

Run the summarize CLI with --extract to fetch the raw content without LLM processing:

summarize "<url_or_path>" --extract --format md --plain 2>/dev/null

This returns the raw markdown content of the source. Save it for analysis.

If the command fails, try with --firecrawl always as fallback (some sites block default extraction):

summarize "<url_or_path>" --extract --format md --plain --firecrawl always 2>/dev/null

If both fail, report the error and stop.

Step 2 — Generate French summary

Run summarize with --cli Codex to generate a French summary:

summarize "<url_or_path>" --lang fr --cli Codex --plain --length long 2>/dev/null

Save the summary output.

Step 3 — Analyze and generate metadata

From the raw content and summary, generate:

1. Title: a clear, concise French title for the resource. If the source has an obvious title (HTML title, PDF title, video title), use it translated to French if needed. Keep it descriptive.

2. Slug: lowercase-kebab-case version of the title for the filename.

3. Tags: merge two sources:

   • User-provided #tags (strip the #)
   • Auto-generated tags (3-8 tags) based on content analysis. Use lowercase, hyphenated tags. Prefer existing vault tag conventions when possible: go, tech, devops, infra, finance, ai, business, health, productivity, etc.

4. Source type: detect from URL/content — article, video, podcast, pdf, documentation, tutorial, talk

5. Language of source: detect the original language (fr, en, etc.)

Step 4 — Scan vault for wikilinks

Before writing, scan vault entities for wikilink enrichment:

• Projects/*.md — project entities + aliases from frontmatter
• Areas/People/*.md — person entities + aliases
• Areas/Companies/*.md — company entities + aliases

Replace mentions of known entities in the summary text with [[Entity-Name]] wikilinks. For alias matches, use [[Entity-Name|alias]].

Step 5 — Create the note

Write to: Resources/ImportedURLs/<Slug>.md

Use this template:

---
type: importedURL
title: "<title>"
source: "<original_url_or_path>"
source_type: <article|video|podcast|pdf|documentation|tutorial|talk>
source_lang: <fr|en|...>
tags: [<tag1>, <tag2>, ...]
status: draft
created: <YYYY-MM-DD>
---

## Resume

<French summary — this is the main content. Well-structured with headers if the summary is long enough. Use ## subsections within the summary for clarity.>

## Points cles

<Bullet list of 5-10 key takeaways from the content, in French.>

## Source

- URL : <source_url>
- Type : <source_type>
- Langue originale : <source_lang>
- Importe le : <YYYY-MM-DD>

If --raw flag was passed, add an additional section before ## Source:

## Contenu original

<raw extracted content, as-is>

Step 6 — Confirm

Report to the user:

Ressource importee : Resources/ImportedURLs/<Slug>.md
Titre : <title>
Tags : #tag1 #tag2 #tag3
Points cles : <count>

Rules

• Always run summarize --extract first, then summarize --lang fr --cli Codex for the summary. Two separate calls.
• If the raw content is very short (< 500 chars), skip the summary step and store the raw content directly with a note that the source was too short for summarization.
• The summary language is always French, regardless of the source language.
• Tags in frontmatter use the YAML array format: [tag1, tag2, tag3]
• File naming: Resources/ImportedURLs/<Slug>.md where Slug is PascalCase-With-Hyphens (matching vault convention, e.g., Migration-4GL-Vers-Go.md)
• Do not duplicate: before creating, check if a file with the same source URL already exists in Resources/ImportedURLs/. If it does, ask the user if they want to overwrite.
• Content language for all note text, section headers, and descriptions is French.
• Keep summaries substantive — aim for 500-1500 words depending on source length. The goal is to capture enough that the user rarely needs to go back to the original source.
• Wikilink enrichment should be subtle: only link entities that are genuinely referenced, don't force connections.

Post-Import — Log

After creating the note, append one line to Areas/Codex/operations-log.md:

| YYYY-MM-DD HH:MM | ingest | import | 1 créée | Resources/ImportedURLs/<slug> |

Create the file with header if it doesn't exist.

Examples

Input: /import https://go.dev/blog/go1.22 Output: Resources/ImportedURLs/Go-1.22-Nouveautes.md with French summary, tags [go, release, programming]

Input: /import https://youtube.com/watch?v=abc123 #ai #talk Output: Resources/ImportedURLs/Conference-IA-Titre.md with French summary of the video transcript, tags [ai, talk, ...]

Input: /import ~/Downloads/rapport-annuel.pdf #finance Output: Resources/ImportedURLs/Rapport-Annuel-Titre.md with French summary, tags [finance, ...]

Input: /import https://example.com/article --raw Output: Same as above but with an additional ## Contenu original section containing the raw extracted text.