create-svg-image
Wrap a raster image in an SVG and overlay a numbered red "step" circle for course screenshots. Use when running /create-svg-image, or when asked to add a numbered callout badge to an image (e.g. a red circle with a "1" pointing at a UI control) that must still render on a GitHub README.
- License 许可
- MIT
- 最近更新
- 2026-11-12
- Stars
- ★ 165
- Forks
- 196
01 前言 Front Matter
- name
- create-svg-image
- description
- Wrap a raster image in an SVG and overlay a numbered red "step" circle for course screenshots. Use when running /create-svg-image, or when asked to add a numbered callout badge to an image (e.g. a red circle with a "1" pointing at a UI control) that must still render on a GitHub README.
02 资源文件 1
- 脚本sample_codes/create-svg-badge.sh9.0 KB
03 SKILL.md 正文
Create SVG Image (Numbered Step Badge)
Use this skill to take a screenshot or image and produce an SVG that shows the image with a numbered red circle placed on top of it. This is for adding "step 1 / step 2" call-out badges that point at a specific UI control in a course screenshot.
The caller always provides three things:
- The image to use — a
.webp,.png, or.jpgfile in the repo. - The location of the numbered circle — where on the image the circle goes, in the image's native pixel coordinates (X from the left, Y from the top).
- The number to apply — the digit/short text shown inside the circle.
Optional inputs: circle --radius, fill --color, --text-color, and --out
path.
Why This Produces an SVG (Not Inline HTML or a Plain Image Edit)
GitHub renders an SVG referenced from Markdown () inside a
sandboxed <img>. In that mode GitHub:
- strips inline
<svg>from README HTML, so you cannot paste raw<svg>markup into the Markdown, and - refuses to fetch an external/relative
<image href="photo.webp">, so an SVG that merely points at the raster shows a blank photo.
The reliable approach is therefore an SVG file that embeds the raster as a
data: base64 URI and draws the <circle> + <text> on top. The photo renders
on GitHub and the badge stays editable as plain SVG markup. (Baking the number
into the PNG/webp with an image editor also works, but then the number is no
longer editable text — prefer the SVG.)
Helper Script
sample_codes/create-svg-badge.sh does the
whole job: detects the image size and mime type, base64-encodes the raster, writes
the SVG, validates it with xmllint when available, and renders a PNG preview
when a renderer is available and previews are enabled.
# Required: --image, --x, --y, --number
bash .github/skills/create-svg-image/sample_codes/create-svg-badge.sh \
--image 02-sessions-worktrees-context/assets/app-create-from-icon.webp \
--x 710 --y 72 --number 1
# -> writes assets/app-create-from-icon-step1.svg and prints a PREVIEW png path
Useful options: --radius (defaults to ~2.5% of the image width so badges stay a
consistent on-screen size, see below), --color "#d1242f" (GitHub danger red),
--out PATH, --grid (coordinate finder, below), --no-preview.
Workflow
-
Find the circle coordinates if the caller gave an approximate spot ("left of the Create from icon") rather than exact pixels. Run the script in
--gridmode, which overlays a labeled coordinate grid on the image and renders a preview without writing anything to the repo:bash .github/skills/create-svg-image/sample_codes/create-svg-badge.sh \ --image <path> --number 1 --gridView the printed
GRID_PREVIEWPNG, read the green (X) and magenta (Y) labels to pick the target pixel, then continue. Coordinates are in the image's native pixel space (the SVGviewBoxequals the image's pixel dimensions). -
Generate the badged SVG with the chosen
--x,--y, and--number:bash .github/skills/create-svg-image/sample_codes/create-svg-badge.sh \ --image <path> --x <N> --y <N> --number <N> -
Verify by viewing the preview. Open the printed
PREVIEWPNG with the image view tool and confirm the circle sits where the caller wanted. Note: macOSqlmanagepads the preview to a square canvas — only the top strip (the image's real height) is the actual image; the white area below is padding, not output. -
Adjust if needed. Nudge by re-running with new values: smaller
--xmoves the circle left, larger--xright;--ymoves it up/down;--radiusresizes it (the font scales automatically). Iterate until the placement is right. -
Wire it into the chapter (only when the caller asks for the README change): reference the new
.svginstead of the original raster, keeping descriptive alt text:
Conventions
- Default output lives next to the source image as
<stem>-step<number>.svg, matching this repo's per-chapterassets/layout. - Keep the original raster in the repo as the editable source; it stays the
input even though the README now points at the
.svg. - Default badge style: solid GitHub danger red
#d1242f, white number, thin white outline so it reads on light or dark screenshots. - Keep the badge diameter consistent across every generated SVG. The radius
defaults to ~2.5% of the image width (the white outline and number scale with
it). Because GitHub scales each screenshot to the README column width, this makes
every badge render at the same apparent size regardless of the capture's
resolution. This is why a
radius 24badge on a 936px-wide capture and aradius 56badge on a 2240px-wide capture look identical. Only pass--radiusfor a deliberate one-off; otherwise let the default keep things uniform. - One circle per file matches the required inputs (image + one location + one
number). For a multi-step figure, either generate separate
-step1,-step2files or hand-add extra<circle>/<text>pairs to the SVG.
Safety and Notes
- Renders previews to a temp dir (
$TMPDIR/create-svg-badge), never into the repo. --gridwrites nothing to the repo; it is purely a coordinate-finding preview.- Screenshots may contain account names, private repo names, or paths — describe the UI generically and don't surface private data when reporting placement.
Related Skills
- comic-strip-site — builds a full webcomic-style site by wrapping
AI-generated character panels in inline SVG (comic frame, tailed speech
bubbles, terminal command bar). Shares this skill's SVG-over-raster technique
and the
--gridcoordinate-finding workflow, but for storytelling panels and recurring cartoon characters rather than single numbered step badges.
04 同仓库更多 Skills
github-copilot-app-automation
Automate and research the GitHub Copilot app desktop experience. Use for driving Copilot App sessions, creating isolated app experiments, mapping app UI/accessibility controls, taking app screenshots, saving PNG captures, optimizing WebP images, testing prompts without submitting, using app/session APIs, canvases, automations, macOS Accessibility, AppleScript, and safe end-to-end Copilot App workflow probes.
check-screenshots-content
Verify GitHub Copilot app screenshots match the course content. Use when running /check-screenshots-content, or when asked to confirm that app- prefixed screenshots in chapter assets folders are consistent with the README labels, button names, menu items, dialog options, and step-by-step instructions that reference them.
book-app-reviewer
Review changes in samples/book-app-web for accessibility, responsive layout, tests/build validation, and small beginner-safe changes.