Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →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.
.claude/skills/davila7-doc/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-10 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -1% | 0% |
soffice and pdftoppm are available, convert DOCX -> PDF -> PNGs.scripts/render_docx.py (requires pdf2image and Poppler).python-docx for edits and structured creation (headings, styles, tables, lists).python-docx as a fallback and call out layout risk.tmp/docs/ for intermediate files; delete when done.output/doc/ when working in this repo.Prefer uv for dependency management.
Python packages:
uv pip install python-docx pdf2imageIf uv is unavailable:
python3 -m pip install python-docx pdf2imageSystem tools (for rendering):
# macOS (Homebrew)
brew install libreoffice poppler
# Ubuntu/Debian
sudo apt-get install -y libreoffice poppler-utilsIf installation isn't possible in this environment, tell the user which dependency is missing and how to install it locally.
No required environment variables.
DOCX -> PDF:
soffice -env:UserInstallation=file:///tmp/lo_profile_$$ --headless --convert-to pdf --outdir $OUTDIR $INPUT_DOCXPDF -> PNGs:
pdftoppm -png $OUTDIR/$BASENAME.pdf $OUTDIR/$BASENAMEBundled helper:
python3 scripts/render_docx.py /path/to/file.docx --output_dir /tmp/docx_pages| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-10 | fail→pass | 10,285 | 2,475 | -76% | 1 | 1 | 0% | 1,838 | 1,043 | -43% | 0 | 0 | — |
case-11 | fail→pass | 13,667 | 4,778 | -65% | 1 | 1 | 0% | 2,618 | 1,484 | -43% | 0 | 0 | — |
case-21 | fail→pass | 16,928 | 7,671 | -55% | 1 | 1 | 0% | 2,572 | 1,732 | -33% | 0 | 0 | — |
case-01 | fail→fail | 28,347 | 6,079 | -79% | 1 | 1 | 0% | 6,203 | 867 | -86% | 0 | 0 | — |
case-02 | fail→fail | 4,667 | 5,590 | +20% | 1 | 1 | 0% | 515 | 873 | +70% | 0 | 0 | — |
case-03 | fail→fail | 6,305 | 4,863 | -23% | 1 | 1 | 0% | 215 | 853 | +297% | 0 | 0 | — |
case-04 | pass→pass | 7,385 | 4,910 | -34% | 1 | 1 | 0% | 1,209 | 1,482 | +23% | 0 | 0 | — |
case-05 | pass→pass | 3,069 | 2,296 | -25% | 1 | 1 | 0% | 599 | 1,056 | +76% | 0 | 0 | — |
case-06 | fail→pass | 12,911 | 7,853 | -39% | 1 | 1 | 0% | 2,346 | 1,987 | -15% | 0 | 0 | — |
case-07 | pass→pass | 5,442 | 3,077 | -43% | 1 | 1 | 0% | 993 | 1,053 | +6% | 0 | 0 | — |
case-08 | fail→pass | 10,963 | 7,950 | -27% | 1 | 1 | 0% | 2,052 | 2,022 | -1% | 0 | 0 | — |
case-09 | fail→pass | 8,891 | 3,727 | -58% | 1 | 1 | 0% | 1,711 | 1,385 | -19% | 0 | 0 | — |
case-12 | fail→pass | 26,294 | 3,033 | -88% | 1 | 1 | 0% | 1,385 | 1,133 | -18% | 0 | 0 | — |
case-13 | pass→pass | 7,444 | 4,335 | -42% | 1 | 1 | 0% | 1,359 | 1,487 | +9% | 0 | 0 | — |
case-14 | fail→pass | 9,575 | 3,662 | -62% | 1 | 1 | 0% | 1,482 | 1,366 | -8% | 0 | 0 | — |
case-15 | pass→pass | 9,870 | 2,111 | -79% | 1 | 1 | 0% | 1,519 | 1,027 | -32% | 0 | 0 | — |
case-16 | pass→pass | 9,671 | 1,684 | -83% | 1 | 1 | 0% | 1,440 | 919 | -36% | 0 | 0 | — |
case-17 | pass→pass | 12,160 | 7,688 | -37% | 1 | 1 | 0% | 2,128 | 1,957 | -8% | 0 | 0 | — |
case-18 | pass→pass | 12,730 | 7,175 | -44% | 1 | 1 | 0% | 2,089 | 2,042 | -2% | 0 | 0 | — |
case-19 | pass→fail | 16,726 | 11,159 | -33% | 1 | 1 | 0% | 2,837 | 2,712 | -4% | 0 | 0 | — |
case-20 | fail→pass | 8,090 | 5,079 | -37% | 1 | 1 | 0% | 1,306 | 1,482 | +13% | 0 | 0 | — |
case-22 | pass→pass | 2,555 | 2,155 | -16% | 1 | 1 | 0% | 365 | 945 | +159% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 18 counted toward the lift figure. The other 4 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +36 percentage points is the difference between those two pass rates over the 18 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.