Doc Cleaner

Documents

doc-cleaner

Doc Cleaner is an Agent Skill for documents published in the notoriouslab/doc-cleaner repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Convert PDF, DOCX, XLSX, and text files to clean, structured Markdown. CJK-friendly, table-friendly, privacy-first.

View on GitHub ↗

QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

Open your project in Codex.
Copy and paste the prompt below.
Review files and risks before approving any action.

Prompt to paste

Inspect the Agent Skill at https://github.com/notoriouslab/doc-cleaner/blob/HEAD/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/doc-cleaner, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

doc-cleaner

Convert documents (PDF, DOCX, XLSX, TXT) to clean, structured Markdown.

When to use

User asks to convert a document to Markdown
User wants to extract text or tables from PDF/DOCX/XLSX files
User wants to clean up bank statements or financial documents
User asks to process a batch of documents in a directory

Commands

Convert a single file (no AI, fastest)

python3 {baseDir}/cleaner.py --input "{{file_path}}" --ai none

Convert a single file with AI structuring

python3 {baseDir}/cleaner.py --input "{{file_path}}" --ai gemini

Convert a single file with Groq structuring

python3 {baseDir}/cleaner.py --input "{{file_path}}" --ai groq

Convert all files in a directory

python3 {baseDir}/cleaner.py --input "{{directory}}" --ai none --output-dir "{{output_dir}}"

Preview without writing (dry run)

python3 {baseDir}/cleaner.py --input "{{file_path}}" --dry-run --verbose

Get machine-readable result summary

python3 {baseDir}/cleaner.py --input "{{file_path}}" --ai none --summary

The --summary flag prints a JSON summary to stdout after processing:

{"version":"1.0.0","total":3,"success":2,"failed":1,"files":[{"file":"report.pdf","output":"./output/report.md","status":"ok"},{"file":"scan.pdf","output":null,"status":"no_content"},{"file":"data.xlsx","output":"./output/data.md","status":"ok"}]}

Options

Flag	Description
`--input, -i`	File or directory to process (required, non-recursive)
`--output-dir, -o`	Output directory (default: ./output)
`--ai`	`gemini`, `groq`, `ollama`, or `none` (default: from config or gemini)
`--password`	PDF decryption password
`--config`	Path to config JSON
`--summary`	Print JSON summary to stdout after processing
`--dry-run`	Preview without writing files
`--verbose`	Enable debug logging

Supported formats

PDF (native, scanned, encrypted), DOCX, XLSX, XLS, CSV, TXT, MD

Exit codes

Code	Meaning
0	All files processed successfully
1	Some files failed (partial success)
2	No processable files found or config error

Notes

Output defaults to ./output/ relative to current directory
For scanned PDFs, AI mode (gemini, groq, or ollama) gives much better results
--ai none requires zero API keys and zero network access
CJK encoding (Big5, CP950, UTF-16) is auto-detected
Tables in DOCX and XLSX are preserved as Markdown pipe tables