process-image
Skill5 starsChanged 8 days agoMITView on GitHub
Resize, convert, rotate, flip, color-adjust, or inspect a local image file (JPEG/PNG/WebP/AVIF, or an SVG rendered to pixels) and make blur-up placeholders — thumbnails, avatars, hero images, fitting an image into a layout, converting between formats, orientation fixes, brightness/saturation tweaks, checking dimensions/size. Runs entirely offline via Bun's native image API, no network call and no fal.ai credit spent. Use whenever the user wants to resize, shrink, convert, re-encode, compress, rotate, mirror, brighten/desaturate an image they already have, needs an LQIP/blur-up placeholder, or needs a PNG/favicon/preview from an SVG logo — even if they just paste a path and say "make this smaller".
In this repository4 other files it gives its agents.
- generate-imageSkill
- generate-svgSkill
- generate-videoSkill
- process-videoSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 15
If the shell says bun isn't found, Bun isn't installed.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
process-image
Wraps Bun's native Bun.Image API (statically linked libjpeg-turbo/spng/libwebp plus OS codecs — this is the "sharp" people mean when they say Bun has it built in, not the npm sharp package) behind one CLI: ${CLAUDE_SKILL_DIR}/scripts/process-image.ts. Purely local — no network, no API key, safe to run freely.
bun ${CLAUDE_SKILL_DIR}/scripts/process-image.ts <command> ...If the shell says bun isn't found, Bun isn't installed. Stop and ask the user to install it from
https://bun.com/docs/installation and restart Claude Code. Don't replace the script with other tools or
code of your own: the checks it runs before a paid call and on the result would be lost.
If you're about to generate a new image and just want it encoded a particular way, don't chain generate → process-image convert/resize — the generate-image generate/edit commands already take --format / --output-quality / --max-width and do it in one step. Reach for process-image for images that already exist on disk.
An .svg source works with every command: it is rendered to pixels first (with resvg, installed by
Bun on first use). resize renders it straight at the target width, so
resize --source logo.svg --output favicon-64.png --width 64 gives a sharp favicon. Render an SVG
this way to look at it before showing it to the user. A preview isn't a deliverable, so it goes in
the project's work folder: media-generation/ inside the project's own scratch folder if it has
one (.scratch/, tmp/), otherwise .media-generation/ holding a .gitignore that contains *,
so git never sees it. Never /tmp or a folder next to the project.
For rotate, flip and adjust, the output format is inferred from the --output extension
(.jpg/.jpeg, .png, .webp, .avif, .heic; anything else is rejected); --quality 1-100 (default 85) applies to any lossy format.
info — inspect before deciding
process-image.ts info --source <path>Prints format, dimensions, and file size. Run this first when you don't already know the source image's size — e.g. before picking a resize width or deciding whether a convert is even needed.
resize
process-image.ts resize --source <path> --output <path> --width N [--height N] [--fit inside|fill] [--quality 1-100]--fit inside (the default) preserves aspect ratio, shrinking to fit within the given width/height. --fit fill stretches to the exact dimensions, distorting the aspect ratio — only use it when the caller explicitly wants an exact box (e.g. a fixed avatar frame). Omit --height to derive it from the width at the source's aspect ratio.
convert
process-image.ts convert --source <path> --output <path> --format jpeg|png|webp|avif|heic [--quality 1-100]webp is the safe default for anything web-bound — good size/quality, transparency, universal support. avif is typically another 20–30% smaller and every current browser renders it, but Bun encodes it through the OS codec: impossible on Linux (nothing to install), on macOS only on Apple M3 or newer, on Windows only with the AV1 Video Extension. HEIC works on macOS and on Windows with the HEIF extension, never on Linux; reading AVIF/HEIC/TIFF sources has the same limits. The script says so when the machine can't do it — then just use webp. --format must agree with the --output extension. png is lossless with transparency (large), jpeg is smallest for photos but has no alpha, heic is for Apple-ecosystem deliverables, not the web. --quality (default 85) affects all lossy formats.
rotate / flip — orientation fixes
process-image.ts rotate --source <path> --output <path> --degrees 90|180|270
process-image.ts flip --source <path> --output <path> [--axis horizontal|vertical]rotate turns clockwise. flip --axis horizontal (the default) mirrors left↔right; vertical turns it upside down. Mirroring is not safe for images containing text or recognizable logos — they'll read backwards.
adjust — brightness / saturation
process-image.ts adjust --source <path> --output <path> [--brightness N] [--saturation N]--brightness and --saturation are multipliers (1.0 = unchanged, 0.5 = half, 1.3 = +30%). There is no hue rotation — Bun's image API doesn't support it. Typical uses: darken a hero image so overlay text stays readable (--brightness 0.7), desaturate a busy background (--saturation 0.6). At least one flag is required.
placeholder — blur-up LQIP
process-image.ts placeholder --source <path>Prints a tiny blurred data:image/png;base64,… URI (a few hundred bytes) to stdout. Put it in the <img>'s CSS background or a low-quality src that swaps to the real image on load — the standard blur-up loading pattern. Nothing is written to disk.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

