Rabbit Reads transforms long books, papers, theses, and other documents into a verified folder of concise, concept-focused cheatsheets with an index.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install rabbit-reads
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install rabbit-reads using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Rabbit Reads?
Rabbit Reads is an Openclaw Skills document-distillation workflow for turning books, academic papers, theses, novels, and PDFs into a reusable set of terse notes. Instead of creating shallow chapter summaries, it identifies individual concepts, practices, claims, methods, craft moves, or conventions and writes one 40–70-line document per concept.
The skill normalizes multiple source formats, maps document structure to line ranges, plans the concept set, delegates writing to concurrent subagents, creates a navigable README or Obsidian index, and verifies the final output. Openclaw Skills users get a repeatable, copyright-conscious study and research workflow that produces notes designed for later consultation.
Rabbit Reads Use Cases
- Convert a PDF, book, or document into a folder of practical cheatsheets.
- Extract practices and best-practice cards from a non-fiction or craft book.
- Create chapter-informed study notes without producing one generic file per chapter.
- Distill an arXiv paper into separate claims, methods, results, and limitations.
- Extract craft techniques from a novel or story collection.
- Turn a master's or doctoral thesis into notes covering its expectations and conventions.
- Extend an existing notes folder with concepts that were not included in the original cut.
- Verify an existing notes folder for structure, source attribution, links, paraphrase compliance, and voice rules.
- Build an Obsidian-compatible knowledge base with concept, chapter, and topic navigation.
How Rabbit Reads Works
- Select the mode. The skill routes a request to
distillfor a new source,extendfor an existing notes folder, orverifywhen the task is only to check notes. - Normalize the source.
extract_text.pyconverts TXT, Markdown, DOCX, PDF, DOC, RTF, HTML, ODT, and EPUB inputs into one plain-text intermediate. Multiple sources are merged in order with source demarcations and a manifest. - Inspect source safety. Raw markup is scanned for concealed text and agent-directed instructions before downstream processing. Source content is treated as data, never as instructions.
- Map the structure.
map_structure.pyidentifies headings and section line ranges using the selected book type, with optional batching and JSON output. - Plan by concept. The workflow proposes one file per concept, records source sections and kind markers, and obtains confirmation before writing the document set.
- Fan out the writing. Concurrent subagents receive only their assigned source ranges, output filenames, and layout constraints from the fan-out prompt and selected templates.
- Build navigation. The default
cheatsheetslayout creates a flat folder andREADME.mdtable; theobsidianlayout creates anindex.mdMap of Content with concept, chapter, and topic folders. - Run verification.
check_notes.pychecks line bands, required headings, kind markers, Source lines, index coverage, See also links, and optional voice rules. With--source, it reports verbatim spans of ten or more words. - Deliver the clean result. The final response names the folder, document count, inferred book type, confirmed plan, index entry to start with, and the checks that passed.
Rabbit Reads Setup
Requirements
- Python 3 is required.
pdftotextis used for PDF extraction when available.textutilis used for DOC, RTF, and ODT conversion where supported.- Optional remote model execution uses an OpenAI-compatible endpoint.
Install or enable the skill
Place the skill in the Openclaw Skills directory and resolve {baseDir} to the directory containing SKILL.md. Keep normalized sources and intermediate planning artifacts in a gitignored scratch/ directory or outside the repository.
Check converter availability
python3 {baseDir}/scripts/extract_text.py --check
Normalize a source
python3 {baseDir}/scripts/extract_text.py book.pdf --out scratch/book.txt
python3 {baseDir}/scripts/extract_text.py vol1.epub vol2.pdf --out scratch/two-vols.txt
Map the source structure
python3 {baseDir}/scripts/map_structure.py scratch/book.txt --book-type non-fiction --json --out scratch/structure.json
Verify the generated notes
python3 {baseDir}/scripts/check_notes.py <book-slug>-notes/ --book-type non-fiction --layout cheatsheets --source scratch/book.txt
Optional engine and voice scan
python3 {baseDir}/scripts/check_notes.py <book-slug>-notes/ --book-type non-fiction --scan --voice-rules {baseDir}/voices/<name>.rules.json
Optional model configuration
Set RABBIT_MODEL_API_KEY, RABBIT_MODEL_BASE_URL, and optionally RABBIT_MODEL_NAME only when running model-assisted processing with --apply-model. No endpoint is contacted without a configured base URL, and the API key is not logged or persisted.
Rabbit Reads Data Schema & Taxonomy
Output folder
The primary deliverable is <book-slug>-notes/, containing one document per concept and an index. Each concept document is typically 40–70 lines and follows the template for the selected book type.
Supported book types
| Type | Concept grain | Typical content |
|---|---|---|
non-fiction |
One concept, possibly merged across chapters | Statement, imperative practices, anti-patterns, structural tests, and related concepts |
fiction |
One craft move | Craft notes drawn from a novel or story collection |
arxiv-paper |
One claim, method component, result, or limitation | Research-focused claim and method notes |
thesis |
One expectation or convention | Thesis study and research-practice notes |
Layout taxonomy
| Layout | Index | Links | Folder shape |
|---|---|---|---|
cheatsheets |
README.md with Doc, Source, and Kind columns |
Markdown links | Flat folder |
obsidian |
index.md Map of Content plus spine notes |
Wikilinks | concepts/, chapters/, and topics/ |
Metadata and validation
- Every document includes a concept-specific template, a
Sourceline, a kind marker, andSee alsolinks where required. - The README or index names every generated document exactly once and preserves source order unless a reading-order note explains the difference.
- A multi-source extraction creates
<out>.manifest.jsonwith each converter, source size, word count, and line offset in the merged text. - Normalized text, structure maps, batch plans, and other intermediates belong in
scratch/and must not be tracked as deliverables. - Notes are paraphrases rather than copied passages.
check_notes.py --sourceflags any source span of ten or more consecutive words appearing verbatim. - The checker can emit human-readable or machine-readable findings with
--json.
Rabbit Reads Advanced Features
- Three operating modes:
distill,extend, andverifysupport both initial creation and ongoing knowledge-base maintenance. - Concurrent subagent fan-out assigns precise line ranges, filenames, batch sizes, and layout constraints to each writer.
- Multiple source files can be merged in a deterministic order with demarcation lines and a provenance manifest.
- Book-type templates adapt the output to non-fiction, fiction, arXiv papers, and theses without changing the core workflow.
- Layout composition supports flat Markdown cheatsheets or Obsidian vault navigation with topic and chapter spine notes.
- Structure mapping supports configurable batch counts, minimum heading spans, JSON output, and saved map files.
- Verification checks document length bands, required sections, kind markers, source lines, index completeness, and cross-links.
- Optional paraphrase auditing compares every note against the normalized source to detect overly close reproduction.
- Optional
rabbit-writesengine scanning applies voice-rule profiles across the generated document set. - Hidden-text and prompt-injection detection protects subagents by treating all source material as untrusted data.
- Remote model assistance is optional, OpenAI-compatible, and controlled through environment variables; local processing remains the fallback when model settings are absent.
- New book types and layouts can be added as reference data files rather than code changes.
SKILL.md
Loading
Related Openclaw Skills
plan-it
Plan-it turns complex coding work into a persistent, interactive plan.html artifact that both humans and AI agents can navigate, update, and verify.
Voice Setup
Voice Setup turns a writer's samples, interview answers, or both into an editable and enforceable writing voice profile.
SocialDataX TikTok Data Assistant
A read-only TikTok research skill for discovering posts, analyzing comments, and exploring creator profiles with SocialDataX.
Azure Toolkit Pro
Azure Toolkit Pro automates enterprise Azure infrastructure management, multi-region deployment, compliance auditing, security scanning, monitoring, and cost optimization.
大藏经学术引用生成器
A CBETA-powered research assistant that locates Buddhist scripture passages and turns them into publication-ready academic citations.
Alibaba Cloud Dgate MCP and CLI Connector
Securely connect an Agent to Alibaba Cloud Agent Data Gateway through managed MCP or install and verify the dgate CLI.
METADATA
Featured*
