doc

Original source / Raw SKILL.md

---
name: "doc"
description: "Use when the task involves reading, creating, or editing `.docx` documents, especially when formatting or layout fidelity matters; prefer `python-docx` plus the bundled `scripts/render_docx.py` for visual checks."
---


# DOCX Skill

## When to use
- Read or review DOCX content where layout matters (tables, diagrams, pagination).
- Create or edit DOCX files with professional formatting.
- Validate visual layout before delivery.

## Workflow
1. Prefer visual review (layout, tables, diagrams).
   - If `soffice` and `pdftoppm` are available, convert DOCX -> PDF -> PNGs.
   - Or use `scripts/render_docx.py` (requires `pdf2image` and Poppler).
   - If these tools are missing, install them or ask the user to review rendered pages locally.
2. Use `python-docx` for edits and structured creation (headings, styles, tables, lists).
3. After each meaningful change, re-render and inspect the pages.
4. If visual review is not possible, extract text with `python-docx` as a fallback and call out layout risk.
5. Keep intermediate outputs organized and clean up after final approval.

## Temp and output conventions
- Use `tmp/docs/` for intermediate files; delete when done.
- Write final artifacts under `output/doc/` when working in this repo.
- Keep filenames stable and descriptive.

## Dependencies (install if missing)
Prefer `uv` for dependency management.

Python packages:
```
uv pip install python-docx pdf2image
```
If `uv` is unavailable:
```
python3 -m pip install python-docx pdf2image
```
System tools (for rendering):
```
# macOS (Homebrew)
brew install libreoffice poppler

# Ubuntu/Debian
sudo apt-get install -y libreoffice poppler-utils
```

If installation isn't possible in this environment, tell the user which dependency is missing and how to install it locally.

## Environment
No required environment variables.

## Rendering commands
DOCX -> PDF:
```
soffice -env:UserInstallation=file:///tmp/lo_profile_$$ --headless --convert-to pdf --outdir $OUTDIR $INPUT_DOCX
```

PDF -> PNGs:
```
pdftoppm -png $OUTDIR/$BASENAME.pdf $OUTDIR/$BASENAME
```

Bundled helper:
```
python3 scripts/render_docx.py /path/to/file.docx --output_dir /tmp/docx_pages
```

## Quality expectations
- Deliver a client-ready document: consistent typography, spacing, margins, and clear hierarchy.
- Avoid formatting defects: clipped/overlapping text, broken tables, unreadable characters, or default-template styling.
- Charts, tables, and visuals must be legible in rendered pages with correct alignment.
- Use ASCII hyphens only. Avoid U+2011 (non-breaking hyphen) and other Unicode dashes.
- Citations and references must be human-readable; never leave tool tokens or placeholder strings.

## Final checks
- Re-render and inspect every page at 100% zoom before final delivery.
- Fix any spacing, alignment, or pagination issues and repeat the render loop.
- Confirm there are no leftovers (temp files, duplicate renders) unless the user asks to keep them.


---

## Referenced Files

> The following files are referenced in this skill and included for context.

### scripts/render_docx.py

```python
import argparse
import os
import re
import subprocess
import tempfile
import xml.etree.ElementTree as ET
from os import makedirs, replace
from os.path import abspath, basename, exists, expanduser, join, splitext
from shutil import which
import sys
from typing import Sequence, cast
from zipfile import ZipFile

from pdf2image import convert_from_path, pdfinfo_from_path

TWIPS_PER_INCH: int = 1440


def ensure_system_tools() -> None:
    missing: list[str] = []
    for tool in ("soffice", "pdftoppm"):
        if which(tool) is None:
            missing.append(tool)
    if missing:
        tools = ", ".join(missing)
        raise RuntimeError(
            f"Missing required system tool(s): {tools}. Install LibreOffice and Poppler, then retry."
        )


def calc_dpi_via_ooxml_docx(input_path: str, max_w_px: int, max_h_px: int) -> int:
    """Calculate DPI from OOXML `word/document.xml` page size (w:pgSz in twips).

    DOCX stores page dimensions in section properties as twips (1/1440 inch).
    We read the first encountered section's page size and compute an isotropic DPI
    that fits within the target max pixel dimensions.
    """
    with ZipFile(input_path, "r") as zf:
        xml = zf.read("word/document.xml")
    root = ET.fromstring(xml)
    ns = {"w": "http://schemas.openxmlformats.org/wordprocessingml/2006/main"}

    # Common placements: w:body/w:sectPr or w:body/w:p/w:pPr/w:sectPr
    sect_pr = root.find(".//w:sectPr", ns)
    if sect_pr is None:
        raise RuntimeError("Section properties not found in document.xml")
    pg_sz = sect_pr.find("w:pgSz", ns)
    if pg_sz is None:
        raise RuntimeError("Page size not found in section properties")

    # Values are in twips
    w_twips_str = pg_sz.get(
        "{http://schemas.openxmlformats.org/wordprocessingml/2006/main}w"
    ) or pg_sz.get("w")
    h_twips_str = pg_sz.get(
        "{http://schemas.openxmlformats.org/wordprocessingml/2006/main}h"
    ) or pg_sz.get("h")

    if not w_twips_str or not h_twips_str:
        raise RuntimeError("Page size attributes missing in pgSz")

    width_in = int(w_twips_str) / TWIPS_PER_INCH
    height_in = int(h_twips_str) / TWIPS_PER_INCH
    if width_in <= 0 or height_in <= 0:
        raise RuntimeError("Invalid page size values in document.xml")
    return round(min(max_w_px / width_in, max_h_px / height_in))


def calc_dpi_via_pdf(input_path: str, max_w_px: int, max_h_px: int) -> int:
    """Convert input to PDF and compute DPI from its page size."""
    with tempfile.TemporaryDirectory(prefix="soffice_profile_") as user_profile:
        with tempfile.TemporaryDirectory(prefix="soffice_convert_") as convert_tmp_dir:
            stem = splitext(basename(input_path))[0]
            pdf_path = convert_to_pdf(input_path, user_profile, convert_tmp_dir, stem)
            if not (pdf_path and exists(pdf_path)):
                raise RuntimeError("Failed to convert input to PDF for DPI computation.")

            info = pdfinfo_from_path(pdf_path)
            size_val = info.get("Page size")
            if not size_val:
                for k, v in info.items():
                    if isinstance(v, str) and "size" in k.lower() and "pts" in v:
                        size_val = v
                        break
            if not isinstance(size_val, str):
                raise RuntimeError("Failed to read PDF page size for DPI computation.")

            m = re.search(r"(\d+)\s*x\s*(\d+)\s*pts", size_val)
            if not m:
                raise RuntimeError("Unrecognized PDF page size format.")
            width_pts = int(m.group(1))
            height_pts = int(m.group(2))
            width_in = width_pts / 72.0
            height_in = height_pts / 72.0
            if width_in <= 0 or height_in <= 0:
                raise RuntimeError("Invalid PDF page size values.")
            return round(min(max_w_px / width_in, max_h_px / height_in))


def run_cmd_no_check(cmd: list[str]) -> None:
    subprocess.run(
        cmd,
        check=False,
        stdout=subprocess.DEVNULL,
        stderr=subprocess.DEVNULL,
        env=os.environ.copy(),
    )


def convert_to_pdf(
    doc_path: str,
    user_profile: str,
    convert_tmp_dir: str,
    stem: str,
) -> str:
    # Try direct DOC(X) -> PDF
    cmd_pdf = [
        "soffice",
        "-env:UserInstallation=file://" + user_profile,
        "--invisible",
        "--headless",
        "--norestore",
        "--convert-to",
        "pdf",
        "--outdir",
        convert_tmp_dir,
        doc_path,
    ]
    run_cmd_no_check(cmd_pdf)

    pdf_path = join(convert_tmp_dir, f"{stem}.pdf")
    if exists(pdf_path):
        return pdf_path

    # Fallback: DOCX -> ODT, then ODT -> PDF
    cmd_odt = [
        "soffice",
        "-env:UserInstallation=file://" + user_profile,
        "--invisible",
        "--headless",
        "--norestore",
        "--convert-to",
        "odt",
        "--outdir",
        convert_tmp_dir,
        doc_path,
    ]
    run_cmd_no_check(cmd_odt)

    odt_path = join(convert_tmp_dir, f"{stem}.odt")

    if exists(odt_path):
        cmd_odt_pdf = [
            "soffice",
            "-env:UserInstallation=file://" + user_profile,
            "--invisible",
            "--headless",
            "--norestore",
            "--convert-to",
            "pdf",
            "--outdir",
            convert_tmp_dir,
            odt_path,
        ]
        run_cmd_no_check(cmd_odt_pdf)
        if exists(pdf_path):
            return pdf_path

    return ""


def rasterize(
    doc_path: str,
    out_dir: str,
    dpi: int,
) -> Sequence[str]:
    """Rasterise DOCX (or similar) to images placed in out_dir and return their paths.

    Images are named as page-<N>.<ext> with pages starting at 1.
    """
    makedirs(out_dir, exist_ok=True)
    doc_path = abspath(doc_path)
    stem = splitext(basename(doc_path))[0]

    # Use a unique user profile to avoid LibreOffice profile lock when running concurrently
    with tempfile.TemporaryDirectory(prefix="soffice_profile_") as user_profile:
        # Write conversion outputs into a temp directory to avoid any IO oddities
        with tempfile.TemporaryDirectory(prefix="soffice_convert_") as convert_tmp_dir:
            pdf_path = convert_to_pdf(
                doc_path,
                user_profile,
                convert_tmp_dir,
                stem,
            )

            if not pdf_path or not exists(pdf_path):
                raise RuntimeError(
                    "Failed to produce PDF for rasterization (direct and ODT fallback)."
                )
            paths_raw = cast(
                list[str],
                convert_from_path(
                    pdf_path,
                    dpi=dpi,
                    fmt="png",
                    thread_count=8,
                    output_folder=out_dir,
                    paths_only=True,
                    output_file="page",
                ),
            )

    # Rename convert_from_path's output format f'page{thread_id:04d}-{page_num:02d}.<ext>' to 'page-<num>.<ext>'
    pages: list[tuple[int, str]] = []
    for src_path in paths_raw:
        base = splitext(basename(src_path))[0]
        page_num_str = base.split("-")[-1]
        page_num = int(page_num_str)
        dst_path = join(out_dir, f"page-{page_num}.png")
        replace(src_path, dst_path)
        pages.append((page_num, dst_path))
    pages.sort(key=lambda t: t[0])
    final_paths = [path for _, path in pages]
    return final_paths


def main() -> None:
    parser = argparse.ArgumentParser(description="Render DOCX-like file to PNG images.")
    parser.add_argument(
        "input_path",
        type=str,
        help="Path to the input DOCX file (or compatible).",
    )
    parser.add_argument(
        "--output_dir",
        type=str,
        default=None,
        help=(
            "Output directory for the rendered images. "
            "Defaults to a folder next to the input named after the input file (without extension)."
        ),
    )
    parser.add_argument(
        "--width",
        type=int,
        default=1600,
        help=(
            "Approximate maximum width in pixels after isotropic scaling (default 1600). "
            "The actual value may exceed slightly."
        ),
    )
    parser.add_argument(
        "--height",
        type=int,
        default=2000,
        help=(
            "Approximate maximum height in pixels after isotropic scaling (default 2000). "
            "The actual value may exceed slightly."
        ),
    )
    parser.add_argument(
        "--dpi",
        type=int,
        default=None,
        help=("Override computed DPI. If provided, skips DOCX/PDF-based DPI calculation."),
    )
    args = parser.parse_args()

    try:
        ensure_system_tools()

        input_path = abspath(expanduser(args.input_path))
        out_dir = (
            abspath(expanduser(args.output_dir)) if args.output_dir else splitext(input_path)[0]
        )

        if args.dpi is not None:
            dpi = int(args.dpi)
        else:
            try:
                if input_path.lower().endswith((".docx", ".docm", ".dotx", ".dotm")):
                    dpi = calc_dpi_via_ooxml_docx(input_path, args.width, args.height)
                else:
                    raise RuntimeError("Skip OOXML DPI; not a DOCX container")
            except Exception:
                dpi = calc_dpi_via_pdf(input_path, args.width, args.height)

        rasterize(input_path, out_dir, dpi)
        print("Pages rendered to " + out_dir)
    except RuntimeError as exc:
        print(f"Error: {exc}", file=sys.stderr)
        raise SystemExit(1)


if __name__ == "__main__":
    main()

```