doc-cleaner skill
doc-cleaner converts PDF, Office, Keynote, Numbers and EPUB into clean Markdown with intact tables and Big5 and CP950 encodings.
- 311 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- In gemini and groq modes document content, including bank statements, goes to a cloud provider
- The script reads and writes files in the directories you point it at
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/doc-cleanerDetects 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.
- In gemini and groq modes document content, including bank statements, goes to a cloud provider
- The script reads and writes files in the directories you point it at
This is third-party code. Review the repository files before installing.
What it does
doc-cleaner reads 16 formats, including PDF, Word, Excel, PowerPoint, Apple Keynote, Numbers and EPUB, and saves each document as Markdown. A PDF is first scanned for tables: found tables become Markdown pipe tables, long text is rebuilt, and scans go to AI recognition mode. The `--ai none` mode extracts text with no keys or cloud. For complex documents Gemini, Groq or a local Ollama is attached. The root SKILL.md gives the agent ready commands, and `--summary` prints the processing result as JSON.
Who it is for. Analysts and developers who need to feed an agent bank statements, reports and books as tidy Markdown without a cloud upload.
Good fit when
- A PDF statement with a multi-column table must become a Markdown table with intact numbers.
- A folder of Word, Excel and Keynote files has to be turned into searchable text at once.
- Documents cannot go to the cloud, and a run with `--ai none` or Ollama fits.
- Scans need an AI recognition pass with cost in mind.
Not a fit when
- Complex PDF layout with formulas and figures must be preserved: text and tables are kept.
- Scans must be recognized without AI keys: they need an AI mode.
- Recursive traversal of nested directories is required: input is processed without recursion.
Example request
Convert `statement.pdf` to Markdown with `cleaner.py` and `--ai none`, then show the transactions table and the processing summary.Limitations
Python 3.9 or newer is required, with dependencies from `requirements.txt`. The `--ai gemini` and `--ai groq` modes send document content to a cloud provider and need a key in `.env`; for private statements choose `--ai none` or Ollama. Cleaning strips typical legal blocks and promotional text from Taiwanese financial institutions, so check important disclaimers against the source. Extended PDF parsing needs Java 11 and extra packages, and scans also need poppler.
How to disable. Delete the cloned `doc-cleaner` directory along with the SKILL.md copy in the agent skills folder. Results are in `./output/` next to the run, and keys are in `.env` and `config.json` inside the project directory.
FAQ
Are keys required?
No, `--ai none` extracts text with no keys or cloud. `GEMINI_API_KEY` or `GROQ_API_KEY` is needed only for AI modes.
How do I process a whole folder?
Pass the directory to `--input`: `python cleaner.py --input ./documents/ --ai none`. Files are processed without entering subfolders, and results appear in `./output/`.
How do I check the result before writing?
Add `--dry-run --verbose`: the script shows what it would do and writes nothing.
Is there a version without Python?
Yes, GitHub releases carry desktop apps: a DMG for macOS and an MSI for Windows with drag and drop.
What about scans?
First run all files with `--ai none`, then run only the scans with `--ai gemini` or another AI mode to avoid paying for extra.
Discussion
No questions or reviews yet.
Related
SkillsA Firecrawl skill and CLI that lets your agent convert Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, CSV and PDF to Markdown on your own machine
A skill that rebuilds PDFs, decks and scans into page-aligned Markdown by checking extracted text against page images.
Docling is a CLI for mixed document archives, bringing PDFs, office files and spreadsheets into a shared format for AI search.
A skill plus the docx program: the agent fills forms, edits tables and reviews Word files with tracked changes
A skill and CLI that turn Markdown into DOCX, PDF, PPTX, XLSX, PNG and about a dozen more formats, all computed locally
The bdc CLI for agents: Word, Excel, PowerPoint and PDF become Markdown, and Markdown with Mermaid becomes a Word document.
README in short
The project targets everyday documents: Chinese encodings are detected automatically, tables are kept, and a separate cleanup removes typical legal inserts in statements.