Slides-Grab Card News Skill
Reuses the slides-grab workflow inside Claude Code / Codex to produce 720×720 square, Instagram-ready card news images.
The skill itself is least-privilege: it only invokes the packaged CLI, explicitly forbids SNS/Instagram publishing automation, produces local PNGs, and describes no network exfiltration or credential operations; however, the install path pulls heavy dependencies (Playwright, sharp), optional image generation involves external providers and reuse of local auth., data-flow disclosure is incomplete, publisher identity is unverified, and rollback is undocumented — so no high score.
Workflow commands are fully consistent with the card-news section of the README (720pt×720pt, card-news mode, default PNG export), and the repo contains card-news-related tests such as editor-card-news-preview-clip.e2e.test.js; but the skill depends on a slides-grab-plan skill not shown in evidence, and error-feedback details for card-news validate/export failures are not demonstrated. Static review cannot exceed the 10-point cap.
The scenario is clear (square Instagram-style card news), boundaries are explicit (local export only, no auto-publishing, experimental formats only on explicit request), and trigger conditions are specific; but non-fit ranges (e.g. no browser environment) are undeclared, there is no Chinese-language support statement, and the core workflow depends on Codex/Claude Code login and npm installation with uncertain mainland-China reachability — hence the deduction.
The SKILL.md has a clear structure (Goal/Workflow/Rules), an explicit MIT license, honest experimental-feature labeling, and stable naming/CLI parameters; but the skill itself lacks versioning, changelog, FAQ, or known-limitations lists, and maintenance responsibility relies on repository-level signals with an unverified publisher — hence mid-high, not full marks.
It claims to reuse the already-workflowed slides-grab pipeline to produce directly Instagram-usable 2160p square PNGs with clear marginal value (automated sizing/export); but a static review cannot confirm actual card-news output quality, the PPTX/Figma paths are self-admittedly unstable, and there is insufficient evidence of verified core-task completion to reach the 7-point cap — hence 6.
Skill claims cross-corroborate with the README and package. test suite (including an editor card-news e2e test), and auditable primary material exists; but nothing was executed, and no third-party execution evidence was provided (the CI workflows shown only cover Pages deployment and npm publishing, with no test CI), so the 5-point cap applies.
- The skill depends on a slides-grab-plan skill not shown in evidence; verify the upstream skill exists and is version-compatible before use.
- PPTX and Figma exports are officially labeled experimental/unstable and may require manual cleanup.
- Optional image-generation paths may reuse local ~/.codex/auth. or require external API keys; review the data flow before enabling.
- The install chain includes heavy dependencies (Playwright, sharp); install from the official npm registry and verify package integrity.
- No Chinese-language support statement; reachability of the core toolchain (npm, Codex/Claude login) from mainland-China networks is unverified.
- Publisher identity is unverified by the FollowSkills registry and treated as unknown; this is a static source review only, with no commands executed.
What does this skill do, and when should you use it?
slides-grab-card-news is a dedicated skill within the slides-grab slide-generation toolkit that converts the standard widescreen presentation workflow into a square card-news pipeline. It reuses the existing plan, design, validation, editing and export stages, but locks every card to 720pt × 720pt and defaults to per-slide PNG export (2160p) because Instagram and most card-news channels consume images, not PDFs. All stages run through the slides-grab CLI in card-news mode. PDF, PPTX and Figma exports are offered only when the user explicitly asks, and the PPTX/Figma paths remain experimental. It deliberately does not automate SNS/Instagram publishing.
Consumes the outline produced by the slides-grab-plan stage; during design and review it pins every card to 720pt × 720pt and runs slides-grab validate, build-viewer, and edit with --mode card-news; it exports by default with slides-grab png --slide-mode card-news --resolution 2160p to emit one square PNG per card (optionally into --output-dir); it only runs slides-grab pdf, convert (PPTX), or figma when explicitly requested, and reminds the user those PPTX/Figma paths are experimental.
- A social media manager turning a topic or document into square image cards ready to post sequentially on an Instagram feed.
- A content creator who wants AI-driven iteration on card copy and layout, with bbox selection in the browser editor for in-place edits.
- A marketing team batch-producing 1:1 card images where the deliverable is PNGs, not a PDF document.
- An existing slides-grab user switching the same deck workflow into card-news mode for social cards.
- Teams who want Playwright-based HTML validation and local-asset checks before exporting cards.
What are this skill's strengths and limitations?
- End-to-end square 720×720 canvas, with output that fits Instagram-style card channels directly.
- Per-card 2160p PNG is the default export, matching how card news is actually distributed.
- Reuses slides-grab's mature validation, single-file viewer, visual editor (bbox selection, direct text editing) and export machinery.
- MIT-licensed, npm-installable, with explicit Claude Code and Codex support.
- PPTX and Figma exports are experimental/unstable — expect layout shifts and manual cleanup.
- PDF, convert, and figma exports are blocked until design-gate records a fresh proceed receipt.
- Depends on the slides-grab CLI and Playwright Chromium runtime; not a pure-documentation skill.
- Explicitly does not implement SNS/Instagram publishing automation — uploads are manual.
- No independent benchmarks or third-party evaluation; evidence is limited to the repo's own docs and demo GIF.
How do you install this skill?
Install the slides-grab collection, which bundles this skill at skills/slides-grab-card-news/SKILL.md. No-clone install: npm install slides-grab, then npx playwright install chromium, then npx slides-grab install-skills --target all --scope user to install the shared Agent Skills plus runtime adapters into Claude Code and Codex locations. Alternatively, after the npm install, add skills via Vercel Agent Skills: npx skills add ./node_modules/slides-grab -g -a codex -a claude-code --yes --copy --full-depth. Requires Node.js >= 20. Clone the repo (npm ci && npx playwright install chromium) only to develop slides-grab itself.
How do you use this skill?
Ask your Claude Code or Codex agent directly, e.g. "Make Instagram-style card news from this outline in card-news mode and export PNGs." The skill then drives four steps: 1) reuse the normal outline process from slides-grab-plan; 2) design every card at 720pt × 720pt and run slides-grab validate / build-viewer / edit with --mode card-news; 3) run slides-grab png --slides-dir <path> --slide-mode card-news --resolution 2160p (defaults to <slides-dir>/out-png, override with --output-dir); 4) export PDF/PPTX/Figma only on explicit request. Note that pdf/convert/figma exports are blocked until a fresh design-gate proceed receipt is recorded, and PPTX/Figma remain experimental and unstable.
How does this skill compare with similar options?
The repository bills itself as an open-source alternative to Claude Design. Compared with the standard widescreen slides-grab skills, this one is specialized for square card news and defaults to PNG rather than PDF; compared with AI tools that generate slide HTML but lack visual editing, slides-grab's differentiator is pointing at a region in the browser editor and asking the agent to edit it in place.