# PDF Text Extract

> Extract text and simple table-like rows from a PDF for downstream AI without OCR binaries. Use when the user asks to read a PDF, turn a PDF into text, pull simple tables from statements/reports, or feed PDF content into an LLM pipeline.

- Skill: `baronguyen001/pdf-text-extract` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add baronguyen001/pdf-text-extract`
- Raw SKILL.md: https://api.skillmd.com/api/skills/baronguyen001/pdf-text-extract/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: DevOps & Infra
- Author: baronguyen001 (https://skillmd.com/u/baronguyen001)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/baronguyen001/pdf-text-extract

---


# PDF Text Extract

Use this skill when a workflow needs machine-readable text from a digital PDF before summarization, classification, or structured extraction. The helper uses pure-Python PDF libraries when available and includes a simple line-based table detector for downstream cleanup.

## When to invoke

- User says: "extract text from this PDF", "feed a PDF to AI", "pull tables from this report", "read this statement".
- The PDF already contains selectable text and does not require OCR.

## When NOT to invoke

- The PDF is a scanned image; use an OCR workflow instead.
- The user needs pixel-perfect table reconstruction with merged cells and layout fidelity.

## Concrete example

User input:

```text
Extract the text and rough tables from this PDF so Gemini can summarize it.
```

Output:

```python
# Copy assets/extract.py into your project, then:
from extract import extract_pdf

doc = extract_pdf("downloads/report.pdf")
print(doc["text"][:2000])
for table in doc["tables"]:
    print(table["page"], table["rows"][:3])
```

Install either `pypdf` or `pdfminer.six` in the target project. No OCR binary is required, and the helper fails clearly when the PDF has no extractable text.

## Pattern to apply

1. Prefer digital text extraction first; do not add OCR unless the source is scanned.
2. Preserve page boundaries so downstream prompts can cite page numbers.
3. Keep table extraction simple: split rows on tabs or repeated spaces, then let a later schema pass normalize columns.
4. Cap text sent to an LLM by page/range when the PDF is long.
5. Fail loudly when no text is found instead of sending an empty prompt downstream.

Reference: `assets/extract.py`.

## Source

Distilled from production use across the author's automation projects. v1.0.0.
See also: [[gemini-structured-output]], [[csv-report-writer]], [[s3-uploader]].

→ Build the full runnable bot with Trawlkit.