Speed Reader

Quickly extract key points from long-form content — articles, PDFs, or URLs — and save structured summaries to your Obsidian vault with source links and reading time estimates.

Overview

Speed Reader helps you process information faster by distilling long content into digestible summaries. It supports three input types (web URLs, PDFs, and pasted text) and three summary formats (1-paragraph, 5-bullets, full digest). Summaries are automatically saved to your Obsidian vault using the PARA method.

Commands

/speed-read <input> [format]

Main command. Summarize content from a URL, PDF path, or pasted text.

Arguments:

• <input> — URL, PDF file path, or quoted text
• [format] — Summary format: paragraph (default), bullets, or digest

Examples:

/speed-read https://example.com/long-article paragraph
/speed-read ~/Documents/report.pdf bullets
/speed-read "Long pasted text content here..." digest

/speed-read-batch <file>

Process multiple inputs from a text file (one per line).

Example:

/speed-read-batch ~/urls-to-read.txt bullets

/speed-read-config

View or update configuration options.

Subcommands:

• show — Display current configuration
• set <key> <value> — Update a setting
• reset — Restore defaults

/speed-read-history

Show recent summaries saved to the vault.

Configuration

Key	Default	Description
default_format	paragraph	Default summary format
vault_path	~/obsidian-vault	Path to Obsidian vault
save_to_vault	true	Auto-save summaries to vault
para_folder	3-Resources	PARA folder for summaries
include_source_link	true	Include source URL/path in note
include_reading_time	true	Estimate and show reading time saved
max_content_length	50000	Max characters to process

Workflow

1. Input Detection

The skill automatically detects input type:

• URL — Starts with http:// or https://, uses browser tools to scrape
• PDF — Ends with .pdf, uses ocr-and-documents skill for extraction
• Text — Quoted or pasted text, processed directly

2. Content Extraction

Web Articles (URLs)

1. Use browser tools to navigate to the URL
2. Extract the main content (strip navigation, ads, footers)
3. Detect article title, author, and publish date
4. Fall back to raw HTML parsing if browser tools unavailable

PDFs

1. Delegate to ocr-and-documents skill for text extraction
2. If PDF is scanned, OCR is applied automatically
3. Extract metadata (title, author, page count)
4. Return extracted text for summarization

Pasted Text

1. Accept text directly from the command
2. Clean formatting artifacts
3. Proceed to summarization

3. Summarization

Based on the selected format:

Paragraph — A single concise paragraph capturing the thesis and key supporting points.

Bullets (5-bullets) — Five bullet points covering:

1. Main thesis/argument
2. Key evidence or data point
3. Notable insight or finding
4. Counter-argument or limitation
5. Takeaway / call to action

Digest — Full structured summary:

# [Title]

**Source:** [URL/path]
**Author:** [Author name]
**Reading Time:** [X] min (Original: [Y] min — saved [Z]%)

## Key Points
- Point 1
- Point 2
- Point 3

## Notable Quotes
> "Exact quote from the text…"

## Analysis
Brief analytical commentary.

## Action Items
- [ ] Follow up on…

4. Saving to Obsidian

Summaries are saved to the vault under 3-Resources/Summaries/:

---
title: "Summary: [Article Title]"
date: 2025-01-15
source: "https://example.com/article"
author: "[Author]"
original_reading_time: 25
summary_reading_time: 2
time_saved_pct: 92
tags: [summary, topic1, topic2]
---

[Summary content here]

Reading Time Estimation

Reading time is calculated at 238 words per minute (average adult reading speed).

Time saved formula:

original_words = word_count(full_content)
summary_words = word_count(summary)
original_time_min = ceil(original_words / 238)
summary_time_min = ceil(summary_words / 238)
time_saved_pct = round(100 * (1 - summary_words / original_words))

Examples

Summarize a web article

User: /speed-read https://arxiv.org/abs/2401.00001 bullets

→ 📄 Fetching article...
→ 📝 Extracting content (3,240 words)
→ ✅ Summary ready (5 bullets)

**Reading time saved: 12 min → 1 min (92% reduction)**

• The paper introduces a novel architecture for…
• Key result: 15% improvement over baseline on…
• The approach uses attention mechanisms combined with…
• Limitation: only tested on English-language datasets
• Takeaway: promising direction for future multimodal research

💾 Saved to: 3-Resources/Summaries/Summary - Novel Architecture.md

Summarize a PDF

User: /speed-read ~/Documents/quarterly-report.pdf digest

→ 📄 Processing PDF (24 pages)...
→ 📝 Extracting text via ocr-and-documents...
→ ✅ Full digest ready

💾 Saved to: 3-Resources/Summaries/Summary - Q3 Report.md

Batch processing

User: /speed-read-batch ~/reading-list.txt bullets

→ Processing 12 items...
→ [1/12] ✅ https://example.com/article-1
→ [2/12] ✅ https://example.com/article-2
→ [3/12] ⚠️ https://example.com/article-3 (paywall — partial content)
→ ...
→ ✅ 11/12 summaries saved to vault

Pitfalls

1. Paywalled content — Browser tools may not bypass paywalls. You'll get partial content with a warning. Consider asking the user for a text paste instead.

2. JavaScript-heavy pages — Single-page apps (SPAs) may not render full content with basic scraping. The browser tools should wait for dynamic content to load, but sometimes content is lazy-loaded.

3. Scanned PDFs — Without OCR, scanned PDFs return no extractable text. Always rely on ocr-and-documents for PDFs, which handles this automatically.

4. Very long documents — For documents exceeding max_content_length, the skill truncates but warns the user. Consider using digest format for long content, which provides the most comprehensive extraction.

5. Rate limiting — Batch processing URLs too quickly can trigger rate limits. The skill adds a 2-second delay between requests by default.

6. Encoding issues — Non-UTF-8 content may produce garbled text. The extraction step attempts encoding detection, but unusual encodings may need manual intervention.

7. Missing metadata — Not all articles have clear author or date metadata. The skill uses fallbacks (OG tags, structured data) but may leave fields blank.

8. Obsidian folder — If the 3-Resources/Summaries/ folder doesn't exist in the vault, it must be created before saving. The skill handles this, but check vault path configuration.

9. Duplicate summaries — Re-summarizing the same URL overwrites the previous summary. The skill uses the source URL as a key for deduplication.

10. Reading time inaccuracy — Content with heavy images, code blocks, or data tables may have skewed reading estimates. The formula counts words only.

Verification Steps

1. Check skill is loaded:

   /speed-read-config show

   Should display current configuration with vault path and default format.

2. Test URL summarization:

   /speed-read https://example.com paragraph

   Should return a concise paragraph summary and confirm save location.

3. Test PDF summarization:

   /speed-read ~/test-document.pdf bullets

   Verify that ocr-and-documents is invoked for PDF extraction.

4. Test text summarization:

   /speed-read "The quick brown fox jumps over the lazy dog. This classic pangram contains every letter of the alphabet." paragraph

   Should return a summary and save to vault.

5. Verify Obsidian save:
   Check ~/obsidian-vault/3-Resources/Summaries/ for the generated markdown files.

6. Verify reading time estimates:
   Confirm that original_reading_time > summary_reading_time and time_saved_pct is displayed.

7. Test batch processing:
   Create a text file with 2-3 URLs and run /speed-read-batch to verify sequential processing.

8. Check error handling:
   Test with an invalid URL, a non-existent PDF, and empty text to confirm graceful error messages.