PixelBin Media Pipeline Skill
Turn chat into a production media pipeline: AI image/video generation, bulk background and watermark removal, free URL transforms, and CDN delivery of e-commerce-ready assets.
Credentials via .env, explicit prohibition on exposing tokens, a 'what NOT to do' list (no scraping, no real-person generation without consent), and confirmation before costly ops. But: assets default to access:'public-read', environments are 'auto-fixed silently', uploads go to a third-party cloud with no rollback/deletion guidance, publisher identity unverified, and data-flow disclosure (images leaving the machine into PixelBin cloud) is limited.
Docs are self-consistent with an error-handling table, retry/concurrency conventions, a cost decision tree, and resumable batch conventions; but scripts/references are not in evidence, there are no test suites or CI evidence, key paths are claimed not reproduced, and failure-feedback quality on abnormal input is unverifiable.
Trigger conditions are clear (keywords, scenario list), boundaries declared, cost-path logic sound; but core function depends entirely on overseas services (pixelbin.io, console/cdn.pixelbin.io) whose reachability from mainland China is questionable, and there is no Chinese-language support disclosure — deducted here.
Well-layered docs (SKILL/INTRO/references/examples), MIT license, .env.example, pinned dependency overrides; but no changelog/versioning, no maintenance-ownership or update-path statement, unverified publisher, and the npm CLI is promised rather than published ('Once published'). Governance is incomplete.
The declared generate→store→transform→deliver workflow is coherent with genuinely useful cost-optimization guidance; but representative outputs are marketing images and sample URLs only, with no independently verifiable end-to-end proof that output is directly usable, and comparative-benefit claims ('zero product drift') are author assertions.
Only in-repo primary material (SDK usage, transform syntax, license); no tests, no CI-run evidence, no third-party verification, and quantitative claims (85+ APIs, 0 vulnerabilities, 17-image showcase) cannot be confirmed in a static read.
- Default public-read publishes uploaded assets; switch to private for sensitive images
- Core function depends entirely on overseas pixelbin.io services; mainland-China reachability and latency undisclosed
- All effectiveness claims are static documentation, unverified; no tests or CI evidence
- Unverified publisher with no maintenance-ownership or update-path commitment
- Quantitative claims (85+ APIs, 0 vulnerabilities) are author statements, independently unverified
What does this skill do, and when should you use it?
This Claude Code skill turns Claude into a full media pipeline built on PixelBin: generate, store (DAM), transform, and deliver images and videos in one flow. It wraps PixelBin's 85+ AI APIs (image generation, video generation, background removal, watermark removal, upscaling, restoration, and more) plus 60+ chainable URL-based image transformations. The skill ships runnable Node.js scripts with concurrency-safe, resumable batch runners. It fits teams that need product imagery processed at scale for e-commerce and marketing.
Accepts images the user pasted, linked, or referenced by local path and saves/uploads them itself; calls pixelbin.predictions.createAndWait to invoke nanoBanana image models and video models such as Sora 2, Veo 3, and Kling 3; uses pixelbin.assets to upload files or URLs and obtain permanent CDN URLs; builds free transformation URLs on cdn.pixelbin.io (e.g. t.resize~t.toFormat~t.compress); runs the scripts under scripts/ (generate-image, generate-video, upload, transform, seo-content, build-page); produces SEO brief. and page-spec. and assembles a landing-page HTML; surfaces pricing upgrade links on quota errors.
- An e-commerce seller with dozens of product photos who needs white-background, 4K, marketplace-sized variants for Amazon/Shopify/Instagram in one prompt
- A content team that wants text-prompt hero images or short videos via Sora 2/Veo 3 with permanent, ready-to-publish CDN links
- An operations person who needs bulk watermark or background removal without touching images one by one
- A site builder who wants SEO content and an AI-illustrated landing page generated from a keyword plus a brand reference site
- A developer who wants to call the bundled scripts/CLI directly for batch jobs of 500–10,000 images
What are this skill's strengths and limitations?
- Full pipeline coverage — generate, store (DAM), transform, deliver — not just one-off API calls
- Explicitly cost-aware design: prefers free URL transforms and reserves expensive generation for when it's actually needed, with a clear decision tree
- Batch runners are concurrency-safe and resumable, suitable for hundreds to tens of thousands of images
- Thoughtful handling of user-supplied images (pasted, URL, local path) and a chat-first UX that hides CLI details by default
- MIT licensed; scripts work standalone without Claude
- Hard dependency on the commercial PixelBin service; generation and AI ops consume metered credits, and video generation is the most expensive operation
- Some AI plugin transforms (e.g. plugin-based erase_bg) require manual activation in the PixelBin console, otherwise the skill must fall back to the predictions API
- Repository has no topics set, and the source provides no test suite or third-party validation
- SEO/landing-page features require a brand reference site or design files; Claude must ask rather than guess the design
- AI prediction outputs are temp URLs (~30-day retention) and must be re-uploaded to the DAM to persist
How do you install this skill?
1) git clone https://github.com/anandpareek-hub/pixelbin-claude-skill.git and cd into the folder; 2) npm install; 3) cp .env.example .env and fill in PIXELBIN_API_TOKEN and PIXELBIN_CLOUD_NAME (both from console.pixelbin.io). Load it one of three ways: A) via Claude Code plugin marketplace — /plugin marketplace add anandpareek-hub/pixelbin-claude-skill then /plugin install pixelbin@pixelbin-claude-skill; B) manual symlink — ln -s "$(pwd)/skills/pixelbin" ~/.claude/skills/pixelbin; C) run scripts directly with node scripts/xxx.js, no Claude needed. The README mentions npx usage once published to npm, but publication status is not evidenced in the source.
How do you use this skill?
After setup, just chat in plain English, e.g.: "I have 12 product photos in ./products/. Generate Amazon-, Shopify-, and Instagram-ready versions for each: white background, 4K upscale, square crop for marketplaces, 9:16 for Instagram Reels. Output a JSON of CDN URLs." The skill orchestrates upload → bg-remove → upscale → multi-aspect transforms → CDN URLs. You can also invoke explicitly with /pixelbin; when running scripts directly, pass job files from examples/.
How does this skill compare with similar options?
The README distinguishes it from generic AI-media skills (single one-off API calls, no CDN, no bulk pipelines): PixelBin provides a full generate → store → transform → deliver stack, free URL-param transforms, and a built-in CDN plus DAM.