slides-grab Export Skill
Reliably convert approved HTML slides to PDF or per-slide PNG, with best-effort PPTX and Figma exports.
The skill requires explicit user approval plus a fresh Stage 2 design-gate Proceed receipt before export; no covert exfiltration; conversion reads local slide HTML and writes local outputs, a least-privilege scope. Deductions: receipt validation logic resides in the packaged CLI and cannot be statically verified, MIT/attribution is clear but publisher is unverified, and no independent security review exists.
Static ceiling is 10. SKILL.md and references are command-consistent (pdf/png/convert/figma/validate); failure-feedback rules are explicit (diagnose root cause, actionable errors, experimental caveats). The repo commits extensive tests for gate blocking and PPTX conversion, but there is no execution evidence and no test-running CI workflow in evidence; abnormal-input behavior is partly 'best-effort', hence deductions.
Scenarios are clear: card-news, widescreen decks, PPTX and Figma import each have explicit commands and trigger conditions; experimental boundaries (PPTX, Figma, text-engine limits) are well declared. Deductions: core function depends on Node>=20, Playwright/Chromium and npm/GitHub downloads, with no Chinese-language support notes and likely reachability friction from mainland-China networks.
Good layering: SKILL.md workflow plus four references (export rules, html2pptx, OOXML, PPTX reference), rich examples, thorough known-limitation disclosure, MIT license. Deductions: no per-skill version or changelog, install/dependency notes live in the README rather than the skill, and maintenance responsibility is only implied via package. author.
Static ceiling is 7. It claims reliable HTML→PDF/PNG conversion and best-effort PPTX/Figma output, with tests covering gate blocking and conversion paths, offering clear marginal value over manual conversion. Deductions: PPTX/Figma outputs are explicitly stated to need manual cleanup, and output quality/representative artifacts are unverified without execution.
Static ceiling is 5. Auditable primary material exists: SKILL.md, references, a committed test suite (design-gate, PPTX conversion tests), and package. dependency manifests that corroborate each other. Deductions: no third-party execution evidence, no CI test workflow in evidence, so conclusions are not independently reproducible.
- PPTX and Figma exports are experimental/unstable; outputs may need manual cleanup and are not one-shot production-grade.
- Export blocking relies on the design-gate Proceed receipt mechanism implemented in the packaged CLI; treat the gate as advisory until independently verified.
- Runtime requires Node>=20, Playwright/Chromium and npm installation; mainland-China networks may face download restrictions — preconfigure mirrors.
- The text engine requires semantic text tags, may reject unsupported HTML/CSS, and rasterizes canvas/SVG visuals.
- Publisher is unverified and this was a static review with nothing executed; validate actual conversion quality with a small trial run.
What does this skill do, and when should you use it?
slides-grab-export is the Stage 3 conversion skill of the slides-grab collection (7 bundled skills), usable in both Claude Code and Codex. It reads approved slide-*.html files and produces PDF, per-slide PNG (including a 1:1 card-news mode), or experimental PPTX/Figma outputs via the slides-grab CLI. Exports are gated: a fresh design-gate Proceed receipt is a hard precondition, and any unresolved Critical finding blocks conversion. PPTX and Figma outputs are explicitly experimental/unstable and may require manual cleanup.
Confirms user approval and a fresh Stage 2 design-gate Proceed receipt; picks the right export target — slides-grab pdf for widescreen decks, slides-grab png --slide-mode card-news --resolution 2160p for card-news/social images; for Chart.js or other canvas charts, runs slides-grab validate to check empty-canvas and opens viewer.html once to confirm final painted state; for PPTX uses slides-grab convert (raster engine for visual fidelity, text engine for editable text) and slides-grab figma for Figma-importable PPTX; on failure, diagnoses the source HTML/CSS and reports actionable errors.
- Teams finalizing agent-generated widescreen HTML decks into distributable PDFs.
- Social-media operators producing Instagram-style 1:1 card news as 2160p PNGs.
- Users who need slides editable in PowerPoint and can accept manual touch-up (text-engine PPTX).
- Designers on Figma Slides who want to import existing HTML decks via PPTX.
- Authors of chart-heavy decks who must verify canvas charts rendered correctly before export.
What are this skill's strengths and limitations?
- PDF and PNG exports are the positioned reliable paths, with up to 2160p/4k resolution.
- The design gate acts as a hard precondition, preventing exports with unresolved Critical findings.
- A clear diagnostic workflow exists for blank Chart.js/canvas charts.
- Works across Claude Code and Codex with documented install paths.
- PPTX and Figma exports are experimental/unstable, with possible layout shifts and manual cleanup.
- The text engine requires semantic text tags, may reject unsupported HTML/CSS, and rasterizes canvas/SVG; --resolution works only with the raster engine.
- Receipt enforcement is strict — stale or missing Proceed receipts hard-block exports.
- No test coverage, performance benchmarks, or failure-rate data are provided in the source.
How do you install this skill?
The skill installs with the slides-grab collection: run npm install slides-grab && npx playwright install chromium, then npx slides-grab install-skills --target all --scope user to install shared skills into Codex and Claude Code; or via Vercel Agent Skills: npx skills add ./node_modules/slides-grab -g -a codex -a claude-code --yes --copy --full-depth. The skill lives at skills/slides-grab-export/ in the repo. Requires Node.js >= 20.
How do you use this skill?
Prerequisite: slides passed the Stage 2 design gate with a fresh Proceed receipt. Example prompts: 'Export decks/my-deck to PDF' or 'Render per-slide PNGs in card-news mode'. Key commands: slides-grab pdf --slides-dir <path> --output <name>.pdf; slides-grab png --slides-dir <path> --output-dir <path>/out-png --resolution 2160p; slides-grab convert --slides-dir <path> --output <name>.pptx --engine raster|text; slides-grab figma --slides-dir <path> --output <name>-figma.pptx. Note that pdf/convert/figma block when the receipt is missing or stale.
How does this skill compare with similar options?
The README positions slides-grab as an open-source alternative to Claude Design; the collection as a whole provides agent-driven editing with visual bbox selection over HTML/CSS slides, while this skill covers only the post-approval export step.