Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Processes Word document files (.docx). Creates, edits, annotates, tracks revisions, analyzes OOXML structure, and preserves formatting for contracts, policies, academic papers, and business documents. Use when working with .docx files or Word documents. Do NOT use for PDFs, spreadsheets, presentations, or plain text files.
.claude/skills/telagod-processing-docx/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -67% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-07 | ✓→✓ | ＝ Same ✓ | -33% | 0% |
.docx is a ZIP archive of XML and resources. Different tasks have different tools and workflows.
| Intent | Workflow | Reference | |--------|----------|-----------| | Read/analyze text only | pandoc → markdown | raw-xml-access.md | | Read structure, comments, media, formatting | unpack → raw XML | raw-xml-access.md | | Create new document | docx-js (JS/TS) | docx-js.md | | Edit own document, simple changes | Document library (Python) | ooxml.md | | Edit someone else's document | Redlining (tracked changes) | redlining.md | | Legal / academic / business / gov docs | Redlining — REQUIRED | redlining.md | | Visual analysis | soffice → PDF → pdftoppm | raw-xml-access.md |
bashpandoc --track-changes=all path-to-file.docx -o output.md # --track-changes=accept/reject/all
docx-js.md (~500 lines). NEVER set range limits.Packer.toBuffer().ooxml.md (~600 lines). NEVER set range limits.python ooxml/scripts/unpack.py <office_file> <output_dir>python ooxml/scripts/pack.py <input_dir> <office_file>See redlining.md for full 6-step workflow with batching strategy and RSID preservation.
| Package | Install | Purpose | |---------|---------|---------| | pandoc | apt install pandoc | Text extraction | | docx | npm i -g docx | Create new docs | | LibreOffice | apt install libreoffice | PDF conversion | | Poppler | apt install poppler-utils | pdftoppm for images | | defusedxml | pip install defusedxml | Secure XML parsing |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-07 | pass→pass | 12,124 | 3,823 | -68% | 1 | 1 | 0% | 1,894 | 1,263 | -33% | 0 | 0 | — |
case-01 | fail→pass | 17,633 | 2,895 | -84% | 1 | 1 | 0% | 3,280 | 1,075 | -67% | 0 | 0 | — |
case-02 | fail→fail | 11,207 | 15,566 | +39% | 1 | 1 | 0% | 1,910 | 3,823 | +100% | 0 | 0 | — |
case-03 | pass→pass | 13,386 | 3,585 | -73% | 1 | 1 | 0% | 2,411 | 1,289 | -47% | 0 | 0 | — |
case-04 | pass→pass | 6,512 | 4,383 | -33% | 1 | 1 | 0% | 968 | 1,344 | +39% | 0 | 0 | — |
case-05 | pass→pass | 15,015 | 9,602 | -36% | 1 | 1 | 0% | 2,320 | 1,964 | -15% | 0 | 0 | — |
case-06 | fail→pass | 11,150 | 4,372 | -61% | 1 | 1 | 0% | 1,955 | 1,395 | -29% | 0 | 0 | — |
case-08 | pass→pass | 13,033 | 8,264 | -37% | 1 | 1 | 0% | 1,927 | 2,070 | +7% | 0 | 0 | — |
case-09 | pass→pass | 9,665 | 5,391 | -44% | 1 | 1 | 0% | 1,368 | 1,571 | +15% | 0 | 0 | — |
case-10 | pass→pass | 17,485 | 7,360 | -58% | 1 | 1 | 0% | 3,129 | 1,707 | -45% | 0 | 0 | — |
case-11 | pass→pass | 10,574 | 3,731 | -65% | 1 | 1 | 0% | 1,796 | 1,186 | -34% | 0 | 0 | — |
case-12 | pass→pass | 7,476 | 2,700 | -64% | 1 | 1 | 0% | 1,200 | 1,060 | -12% | 0 | 0 | — |
case-13 | pass→pass | 11,415 | 4,419 | -61% | 1 | 1 | 0% | 1,875 | 1,352 | -28% | 0 | 0 | — |
case-14 | fail→pass | 8,586 | 5,125 | -40% | 1 | 1 | 0% | 1,542 | 1,335 | -13% | 0 | 0 | — |
case-15 | pass→pass | 9,827 | 2,632 | -73% | 1 | 1 | 0% | 1,567 | 1,044 | -33% | 0 | 0 | — |
case-16 | fail→pass | 13,324 | 2,511 | -81% | 1 | 1 | 0% | 2,139 | 1,031 | -52% | 0 | 0 | — |
case-17 | pass→pass | 14,641 | 4,055 | -72% | 1 | 1 | 0% | 2,156 | 1,328 | -38% | 0 | 0 | — |
case-18 | pass→pass | 10,558 | 5,651 | -46% | 1 | 1 | 0% | 1,360 | 1,461 | +7% | 0 | 0 | — |
case-19 | pass→pass | 4,018 | 1,663 | -59% | 1 | 1 | 0% | 649 | 924 | +42% | 0 | 0 | — |
case-20 | pass→pass | 2,701 | 1,748 | -35% | 1 | 1 | 0% | 440 | 952 | +116% | 0 | 0 | — |
case-21 | pass→pass | 7,063 | 4,872 | -31% | 1 | 1 | 0% | 1,243 | 1,502 | +21% | 0 | 0 | — |
case-22 | pass→pass | 6,914 | 2,875 | -58% | 1 | 1 | 0% | 1,303 | 1,192 | -9% | 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. The headline lift of +18 percentage points is the difference between those two pass rates over the 22 comparable cases.
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.