Skill Seekers
yusufkaraaslan/Skill_SeekersPython tool and MCP server that converts documentation sites, repos, PDFs and other sources into agent skills and RAG-ready knowledge, with 22 export targets.
with caveats
Security: Check first depth3/5
book-to-skill
Converts a technical book or document folder into a structured agent skill with per-chapter files your agent loads only when a question needs them.
recommended
A thorough, well-tested converter that does what it describes; just keep skills made from third-party books private, as the README itself advises.
check first: installs packages · relies on an outside CLI → security details
- Depth
- 5/5
- Last update
- 4 days ago
- Installs
- —
- Stars
- 37K
- Forks
- 3.9K
- License
- MIT
What book-to-skill does
book-to-skill is a single, large skill that turns documents into another skill. You run /book-to-skill ./my-book.pdf, or point it at a folder, a glob or a list of files. The agent then follows a numbered procedure that pulls out named frameworks, principles, techniques, anti-patterns and the author's voice. It does not write a summary. The output goes to ~/.agents/skills/<slug>/. It contains a SKILL.md of about 4,000 tokens with the core mental models and a chapter index, one file per chapter in chapters/ (about 1,000 tokens each), and glossary.md, patterns.md and cheatsheet.md. You then ask something like /my-book replication, and the agent reads the relevant chapter instead of the whole book.
The skill has four modes: full conversion, analyze-only (it stops with an extraction report for you to review), generate from a prior analysis, and an update mode that folds new material into an existing skill. Supported inputs are PDF, EPUB, DOCX, HTML, Markdown, plain text, RTF, and MOBI/AZW through Calibre. Behind the 55 KB SKILL.md are 29 reference files, 85 scripts, 69 test files and CI. One script, discovery_tax.py, compares token cost against dumping the book into context. The README reports 24x to 51x fewer tokens per question, measured on real books.
Good for
- Developers who want a technical book they own available as an on-demand reference while coding
- Turning internal docs or a folder of sources into one queryable skill
- Studying a book through your agent without pasting the whole PDF into context
Not for
- Publishing or sharing skills built from copyrighted books
- Quick one-off questions where pasting a page or two is enough
What's inside depth 5/5
One 55 KB SKILL.md backed by 29 reference files, 85 scripts, 7 templates, 69 test files, 13 eval files and CI.
- Skills
- 1
- Reference docs
- 29
- Scripts
- 85
- Tests / evals
- 82
- Templates
- 7
- Avg SKILL.md
- 53.7 KB
Counted from the repository file tree. How we score depth.
Dependencies and costs
- Python (the helper scripts and tests are Python)
- Calibre, only for MOBI/AZW input
Security review
Security: Check firstThe skill runs a bundled Python extractor locally on your books and documents and writes generated skill folders into your user-level agent skill directories. It may symlink them into ~/.claude/skills. It can pip-install optional parser packages (with a prompt in interactive use) and relies on external tools like Calibre and pdftotext, and no network uploads of your documents appear in the code we reviewed.
Worth knowing
- Installs packages or toolsCan run
pip installfor missing optional parsers (docling, pypdf, pdfminer.six, ebooklib, beautifulsoup4, python-docx, striprtf, trafilatura, pdf-inspector). It prompts first in interactive sessions and falls back to built-in parsers in non-interactive ones unless install mode is 'yes'.book_to_skill/dependencies.py:118 - Relies on a CLI or package we could not reviewCalls Calibre's
ebook-convertbinary for MOBI/AZW files, and poppler'spdftotext/pdfinfofor PDFs (parsers/pdf.py).book_to_skill/parsers/calibre.py:47 - Changes files outside your projectWrites generated skills to user-level skill roots such as ~/.agents/skills and creates a symlink in ~/.claude/skills for Claude Code. It may also run
hermes skills trustfor project roots.SKILL.md
What we checked
The repository (132 files) on Nov 1, 2026. Not a full review: we read 21 of 85 code files; it relies on code outside the repository: Calibre `ebook-convert` binary (used for MOBI/AZW); poppler-utils `pdftotext` / `pdfinfo`; Optional pip packages installed on demand: pdf-inspector, docling, pypdf, pdfminer.six, ebooklib, beautifulsoup4, python-docx, striprtf, trafilatura.
NVIDIA SkillSpector static scan: 108 pattern hits (Dangerous Code Execution 38, Prompt Injection 17, MCP Rug Pull 13, Data Exfiltration 12, Privilege Escalation 10, analysis-evasion 6). The bidi and invisible-Unicode hits are test fixtures for the sanitizer and Persian/Sinhala text in chapter-detection regexes. The 'Ignore all previous instructions' and /etc/passwd strings sit in tests of the injection scanner and output-dir safety, and the os.environ hits are test subprocesses copying the environment. The SKILL.md HTML comment is informational host-compatibility notes.
A review, not a certification. Skills change after we check them — read the files before installing.
License
MIT, commercial use allowed. The README warns that generated skills of copyrighted books should stay private and not be redistributed.
Can it run unattended?
Not really. A conversion is a bounded job that ends once the skill files are written, and analyze-only mode deliberately stops for your review, so there is no long-running loop to leave unattended.
Alternatives to book-to-skill
Anthropic Agent Skills
anthropics/skillsAnthropic's reference skills repo: 20 skills including document tools (docx, pptx, xlsx, pdf), frontend-design, skill-creator and mcp-builder.
recommended
Security: Check first depth4/5
task-observer (One Skill to Rule Them All)
rebelytics/one-skill-to-rule-them-allA meta-skill that watches your sessions, logs corrections and repeated work, and proposes improvements to your other skills for you to approve.
recommended
Security: Check first depth4/5
Engram
nagisanzenin/engramA learning system for the human user: the agent tutors you, a blind subagent grades your free recall, and an FSRS scheduler brings topics back for review.
recommended
Security: Check first depth4/5
FAQ
What is book-to-skill?
It is an open-source agent skill that converts a book or set of documents into a new skill. The generated skill has a core SKILL.md, one file per chapter, a glossary, a patterns file and a cheatsheet, so the agent can answer from the source text by loading only the chapter it needs.
How do I use book-to-skill?
After installing it (the README links to docs/install.md for each agent), run /book-to-skill ./my-book.pdf or pass a folder, glob or list of files. The new skill is written to ~/.agents/skills/<slug>/, and you query it with a command like /my-book replication.
Which file formats does book-to-skill support?
PDF, EPUB, DOCX, HTML, Markdown, plain text and RTF. MOBI and AZW files work if Calibre is installed.
Is book-to-skill safe to install?
The scanner hits we reviewed were false positives, mainly test data for a sanitizer that strips invisible Unicode direction characters. We found no telemetry, credential access or hidden network calls in the material we saw. Its helper scripts run locally when you convert a document.
Can I share a skill I made from a book?
The README says not to redistribute skills built from copyrighted works, because that can infringe the rights holder. It suggests keeping those private. Internal docs, your own writing and openly licensed material are a different case, depending on their license.
book-to-skill vs Skill Seekers: which should I use?
Skill Seekers converts documentation websites, GitHub repositories and PDFs into Claude skills. book-to-skill concentrates on books and document files and produces a chapter-by-chapter layout with a glossary, patterns and cheatsheet. Pick book-to-skill for long-form books you study, Skill Seekers if your source is a docs site or repo.
Sources: GitHub repository (stars, license, file tree, last push Oct 28, 2026). Review based on the README and SKILL.md files as published.