Knowledge Base Kit skill
Skills for Claude Code and Codex turn PDF, DOCX and HTML into a local knowledge base with the original, full text and summary.
- 274 ★ on the repository
- Last commit
- AGPL-3.0
Extended access
Why this level
- Documents and images are sent to the chosen model provider, and scripts write files into the workspace.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/knowledge-base-kitDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Documents and images are sent to the chosen model provider, and scripts write files into the workspace.
This is third-party code. Review the repository files before installing.
What it does
Knowledge Base Kit installs the `process_docs` and `kb` skills for Claude Code and Codex into a separate workspace. The agent takes a document from `inbox/`, extracts text with scripts, describes images in isolated workers and writes `full.md` and `summary.md`. The original attachment stays beside them, and `catalog.md` lets an agent find material without opening large files. Structure, links, archive folders and TeX formulas are validated before filing. This is an agent-assisted workflow rather than a hosted service: the agent drafts the final text and a person reviews it before it enters the base.
Who it is for. Researchers, analysts and teams accumulating PDFs, reports and notes who want an agent to answer questions grounded in sources.
Good fit when
- A folder of PDFs and contracts has grown and an agent needs summaries linked to originals.
- Scientific papers with formulas must become readable Markdown.
- Jupyter notebooks and web pages should be gathered into one searchable catalog.
- A team wants to review results by hand before a document enters the base.
Not a fit when
- Fully automatic ingestion without human review is wanted.
- Documents cannot be sent to a model provider.
- A hosted service with a web interface is required.
Example request
Use the local `process_docs` skill with the Codex backend on `inbox/quickstart.md`, prepare the full document and summary, show the category and changed files and wait for my review.Limitations
Python 3.10+, Node.js 20+, Pandoc 3+ and an authenticated Claude Code or Codex CLI are required. "Local" describes where files live, not local-only inference: images and short context go to the chosen provider, and so do text drafts. Structure checks do not prove factual accuracy or complete coverage. The license is AGPL-3.0, and the PyMuPDF and html2text dependencies are GPL-family.
How to disable. Delete the knowledge-base workspace or the `.kbkit/` directory inside it. The installer does not edit global agent settings, so no further cleanup is needed.
FAQ
Can I try it without an account or model calls?
Yes. `python scripts/offline_demo.py ../kb-demo` extracts the bundled Markdown and validates a reference output; the demo does not simulate AI generation.
Where do my documents live?
In a separate workspace next to the project, not inside the toolkit repository. Never commit a private base or logs to the toolkit repository.
How do I get Chinese summaries?
Add `--language zh-CN` when initializing the workspace.
Which formats are processed?
PDFs, DOCX, static HTML, notebooks, text and images.
Which backends are available?
The `--backend` flag accepts `claude`, `codex` or `both`.
Discussion
No questions or reviews yet.
Related
SkillsThe official Kingsoft Docs (WPS) skill set: Claude Code and Codex create and edit cloud documents through kdocs-cli
A set of three research skills: paper writing, academic Word and PowerPoint documents, scientific computing and figures
The /okf skill builds a Markdown knowledge graph from code and docs in the Open Knowledge Format and keeps it up to date.
Four Claude Code and Codex skills: Word with tracked changes, PDF forms and OCR, PowerPoint from HTML, Excel recalculation.
16 ready skills for Cursor, Claude Code and Codex: theses, PDF and Office, draw.io diagrams, project documentation
Two skills for Claude Code and Codex: the agent translates papers in the EasyRead library and answers your margin notes
README in short
A workspace holds `inbox/`, `catalog.md`, categories and a private `.kbkit/` directory. Each document gets `full.md`, `summary.md` and its original attachment, each category gets `README.md`.