PPTX Presentation Skill
Create, edit, parse, and validate PowerPoint presentations.
The evidence shows local PPTX/POTX processing with safe archive extraction, path checks, and some failure safeguards. However, the skill invokes shell, Python, Node, and LibreOffice; add_slide defaults to in-place rewriting and clean.py can delete unreferenced files. It requires no user confirmation, backup, or uniform rollback plan, and does not explain file data flow, sensitive-content handling, or dependency supply-chain controls. Eight points are deducted. Verified Anthropic provenance supports attribution but does not add safety points.
The instructions and scripts cover creation, editing, validation, rendering, and several abnormal cases in concrete detail, with path, relationship, and archive checks in the code. Still, dependency availability, key-path reproduction, and boundary-input behavior lack static test or CI evidence, while the complex LibreOffice shim path is unexecuted and uncertain. The score is capped by static calibration and set to 8/20.
Trigger semantics are explicit and broad, covering .pptx/.potx and mentions of decks, slides, and presentations, with distinct workflows for creation, editing, reading, and templates. Non-fit ranges, input/output contracts, and Chinese-font/content guidance are limited. The core function does not entirely depend on overseas online services, but local dependency and mainland-China environment fit are unverified, so 5 points are deducted.
The skill has strong information architecture, script tables, dependency notes, pitfalls, known limitations, and QA procedures. Versioning, changelog, maintenance ownership, and update path are unclear; the license is marked Proprietary, but the supplied evidence does not include the complete LICENSE.txt, and installation and troubleshooting retain hidden assumptions. Six points are deducted.
The skill addresses generation, editing, parsing, template work, structural validation, and visual QA, with relatively direct procedures and repair guidance. However, this review has no executed representative outputs or execution logs; users still need environment setup and review. Under static calibration, the score is limited to 6/15.
The supplied SKILL.md and script source provide auditable implementation evidence, and the validation code reports several concrete defect classes. There are no execution logs, real CI workflows, committed tests covering key paths, or third-party reproduction records, so the evidence is mainly a single source-code set and scores 4/10.
- add_slide.py may rewrite the input in place by default; clean.py and example rm commands can remove files, so users should back up and confirm targets first.
- Required dependencies such as pptxgenjs, markitdown, Pillow, defusedxml, lxml, LibreOffice, and pdftoppm are not version-pinned, so environment differences may interrupt the workflow.
- There is no dedicated evidence for Chinese fonts, Chinese layout, or mainland-China software availability; the static documentation should not be treated as a tested production guarantee.
What does this skill do, and when should you use it?
The PPTX skill handles .pptx and .potx files for creation, reading, editing, splitting, combining, and template-based work. It selects approaches such as pptxgenjs scripting, XML editing, or markitdown extraction based on the task. It also provides scripts for thumbnails, slide duplication, cleanup, validation, and visual QA, plus detailed design and chart guidance. The repository describes the document skills as source-available rather than open source.
Writes pptxgenjs scripts for new decks; unpacks and edits PowerPoint XML for existing decks and templates; extracts slide text with markitdown; creates labeled thumbnail grids with thumbnail.py; duplicates slides with add_slide.py; removes unreferenced resources with clean.py; validates structure, relationships, charts, and slides with validate.py; and converts decks through LibreOffice and pdftoppm for visual inspection.
- A team creating a new presentation, pitch deck, or business report.
- A user extracting text from an existing .pptx file.
- A user building a presentation from a .pptx or .potx template.
- A user modifying, reordering, splitting, or combining slides.
- A user checking a deck’s file structure, charts, content, and visual rendering before delivery.
What are this skill's strengths and limitations?
- Covers creation, reading, editing, template handling, and quality assurance.
- Includes scripts for thumbnails, slide duplication, cleanup, and validation.
- Provides concrete rules for charts, typography, layout, color, and visual quality.
- Requires content, file, and visual QA before delivery.
- No standalone package, version information, or test suite is documented in the source.
- Several workflows depend on Node.js, Python, LibreOffice, Poppler, and multiple Python and Node.js components.
- The README describes the document skills as source-available rather than open source; SKILL.md labels the license Proprietary.
- The source does not explain integration details for products outside Claude.
How do you install this skill?
The source documents Claude Code installation with /plugin marketplace add anthropics/skills, followed by Browse and install plugins, anthropic-agent-skills, and document-skills. It also gives /plugin install document-skills@anthropic-agent-skills. The README says these example skills are available to paid Claude.ai plans. No standalone general installation procedure is documented.
How do you use this skill?
Trigger it whenever a task involves a .pptx or .potx file, or mentions a deck, slides, or presentation. For example: Use the PPTX skill to create a presentation about quarterly results. To read content, use markitdown deck.pptx. After generating a deck, run python scripts/office/validate.py output.pptx and follow the documented image-rendering and visual-QA workflow.