With Arguments

Automatically process unsummarized files and generate structured summaries.

With Arguments

• File path: Process single file (e.g., /summarize article.pdf)
• Folder path: Process all files in folder (e.g., /summarize ~/Documents/Articles/)
• URL: Fetch and summarize web page

Input Detection

Automatically detect input type:

• PDF: Use PDF skill for extraction (preferred) or Read tool for < 32MB files
• SRT: Extract text content, ignore timestamps
• MHTML: Decode quoted-printable, extract substantive paragraphs
• TXT: Direct read
• URL: Use WebFetch

Output Format

Generate structured summary with readable spacing:

## **{Type}{Number}. {Title}**
*({Source Type} - {Context})*

📍 **Contexte** - [1 phrase résumant l'enjeu]

📝 **Fait**

**{Sous-section 1}**
- Point clé 1
- Point clé 2
- Point clé 3

**{Sous-section 2}**
- Point clé 1
- Point clé 2

👿 **Détracteur** - [Point de vue opposé ou limite] ([Consensus: majoritaire/minoritaire/à débattre])

🔖 **Action**

- Implication 1
- Implication 2
- Ce qu'il faut surveiller/anticiper

Types: V (vidéo), P (PDF/article), A (autre) Numbering: Séquentiel par type (V1, V2... P1, P2...)

Arguments

Parse $ARGUMENTS:

• No arguments: Process ~/Downloads/Summarize/ (default)
• File/folder path: Process specified location

Processing Workflow

Phase 1: Discovery

1. Determine target (from $ARGUMENTS or default folder)
2. Find or create resume file {YYYYMMDD}_Resume.md
3. List supported files: .pdf, .srt, .mhtml, .txt
4. Identify unsummarized files (not already in resume)

Phase 2: Summarization

For each unsummarized file:

1. Detect input type from extension
2. Extract content
3. Generate structured summary
4. Append to resume with --- separator

Phase 3: Cleanup

Delete processed source files from Pending/ folder after successful summary. Show list of deleted files.

Technical Notes

• Always append, never overwrite existing summaries
• Add horizontal rule --- separator between summaries
• Blank lines between sections for readability
• Bold for sub-sections, numbers, key actors/concepts

Style

• Language: Titles in EN, content matches source language
• Tone: Pragmatic, frugal, actionable
• Emphasis: Bold for key terms/numbers/sub-sections, NOT for entire sentences
• Concision: Dense information per line, but well-spaced overall

Error handling

• Missing dependencies (pdftotext): Install via brew install poppler
• Unsupported format: Inform user, suggest manual processing