Render pixel-accurate ChatGPT mobile (iOS) screen mockups in light mode from a thread JSON. Supports user text bubbles, user image attachments, assistant markdown prose, citation chips, the OpenAI spiral logo, the Apps-SDK GPT chip in the composer, and three header styles (model-tag, plain title, "Get Plus"). Fixed 9:16 viewport. Outputs HTML + PNG.
Generate single-frame ChatGPT mobile screens that read as real iOS captures — accurate header chrome, user/assistant message styling, composer pill, status bar, and the OpenAI spiral mark. Designed for use inside ad creatives, social posts, video storyboards, and product mockups.
If you need an animated, video-recorded chat reveal (bubbles popping over time, SFX, music bed, end card), use the molecule skills/molecules/create-chatgpt-video-ad. This atom only produces a single still frame.
Inputs
Compose a thread.json matching the schema below, then pass it to render.js.
"plain-title" — ≡ Get Plus ✦ ●● — for the upsell/empty-state title.
header.rightIcons — array of icon keys, rendered left-to-right on the right side. Available: "personPlus", "dottedCircle", "edit", "more". Default is ["personPlus", "dottedCircle"] for model-tag, ["edit"] for title-only.
messages[] — one of:
{ "type": "user-image", "src": <path>, "aspect": "square"? } — relative src paths resolve against the thread.json directory.
{ "type": "assistant", "text": <markdown>, "title"?: <string>, "feedback"?: <bool> } — left-aligned plain prose. If title is set, an OpenAI spiral logo + bold title is rendered above the body. feedback: false hides the thumbs-up/down chips.
parent dir for the dated output folder; default ./chatgpt-mockup-exports/
--name <slug>
override the output folder slug
Output
<output>/<YYYY-MM-DD>-<slug>/
index.html # standalone HTML (CSS inlined)
screenshot.png # 2250 × 4002 PNG (DPR 3, 9:16)
thread.json # copy of the input for reproducibility
Workflow
Receive a brief or existing thread JSON.
If only a brief is given, compose thread.json matching the schema above.
Run node render.js --thread <path>.
Open the PNG. If anything looks off (header alignment, bubble width, missing icon), edit the JSON or templates/chat.css and re-render.
Setup (one-time)
cd create-chatgpt-mockup
npm install
npx playwright install chromium
Quality checks
Renders in light mode only (white background, dark text). Dark-mode rendering is out of scope for v1.
Status bar time and right cluster (signal / wifi / battery) sit at the very top, no clipping under the iOS status-bar safe area.
Header chrome icons (hamburger / personPlus / dottedCircle / edit / more) all visually weigh the same as a real ChatGPT iOS screenshot at the same display size — eyeball, don't pixel-match. Sub-reading-weight icons read as broken even when their pixel dimensions "match the spec."
Header model tag (5.1, 5.2) sits inline with ChatGPT and its chevron — no wrapping.
User bubble is right-aligned, soft-gray (#F4F4F4), border-radius: 28px, max-width 78% of stage.
User image attachments use the same 28px corner radius and sit right-aligned above the text bubble.
Assistant body is plain prose (no bubble), left-aligned, full-width within the gutter.
Composer corner radius is 999px (full pill), 1px border #E5E5E5, send button is 44 × 44 round.
When composer.text is set: send button is solid black with a white up-arrow. Caret sits right after the last character of typed text.
When composer.streaming: true: send button shows a black square stop icon.
When a citation marker [[cite:Foo +1]] is in the assistant text, it renders as a pill chip inline at the end of the sentence.
Failure modes
Symptom
Cause
Fix
Thread file not found
Wrong path or running from a different cwd
Use absolute paths or run from the skill directory
Image attachment shows broken-image glyph
src is relative but the renderer couldn't resolve it
render.js resolves relative src against the thread.json's directory — make sure the asset exists at that path
Send button stays light gray after typing
composer.text is empty string "" instead of unset
Either omit text entirely or set it to a real string
Caret renders as a stray | separated from the text
Caret span is being treated as a flex child
The CSS rule for .composer .input uses inline content flow — do not switch it to display: flex
Streamed list-item words render with no spaces ("Howconversationswork")
Use display: inline for .word — modern browsers honor filter: blur() + opacity transitions on inline elements too
Bullets visible while list-item text is still hidden during streaming
CSS list-style markers are browser-painted; the word-stream wrapper only sees text nodes
The atom auto-converts <ul><li>X</li></ul> to <div class="md-list"><div class="md-li">• X</div></div> when stream: true — keep that pass; don't regress it. Any new markdown that emits browser-painted glyphs (::marker, quotes, counter-content) needs the same text-conversion treatment.
Markdown --- doesn't render an <hr>
The --- line isn't on a line by itself, or the surrounding text didn't end its paragraph
Surround --- with blank lines on both sides
Stanzas of a poem joined into single paragraphs
Lines separated by single \n get joined as a single paragraph but with <br> between them — this is the intended behavior. If you want stanza breaks, use \n\n between stanzas.
Header right cluster overlaps the model tag on long titles
Title text is too wide for the centered slot
Shorten header.title, drop the model tag, or switch header.style to "title-only"
Render every example into tests/output/ for visual review
Known limitations
Light mode only. Dark mode would require a separate theme block and is out of scope for v1.
No carousel / horizontal-scroll cards. The Apps-SDK SeatGeek reference shows a horizontal card carousel inside the assistant response; not modeled here.
No live keyboard rendering. The composer is the bottom pill only — no on-screen keyboard chrome, suggestion bar, or QWERTY grid. (If you need the keyboard, scale the screen down 50% and overlay a separate keyboard image.)
Markdown is intentionally minimal — no tables, no fenced code blocks with syntax highlighting, no nested lists. Add to generate.js#renderMarkdown if a brief requires more.
Emoji rendering depends on system fonts. Chromium on macOS ships Apple Color Emoji; on Linux ships Noto Color Emoji.
Quality Checks
Run bash tests/run-all.sh and open the PNGs in tests/output/, then confirm:
Every example renders without error. All six examples/*.json produce an index.html + screenshot.png + thread.json triplet; render.js exits 0.
Dimensions are stable. Each screenshot.png is 2250 × 4872 (the .stage is 750 × 1624 and screenshot.js captures at deviceScaleFactor: 3). A wrong size means the stage box or DPR changed.
Light mode, white background. Background is white, text near-black. This atom does not render dark mode.
Status bar. Time sits top-left; the signal / wifi / battery cluster sits top-right, unclipped. When statusBar.dnd: true (see 02-sleep-apnea-long.json, 05-seatgeek-gpt-chip.json) a moon glyph renders next to the time.
Header style matches the JSON.model-tag shows ChatGPT 5.1 ▾ (01-sunscreen-image.json, 06-short-howto.json); title-only shows the plain title + right chevron (02, 03, 05); plain-title shows the upsell title with no chevron (04-empty-typing.json). The center cluster never wraps and the right icons never overlap it.
User turns. User text is a right-aligned soft-gray bubble (#F4F4F4, 28px radius). A user-image (see 01-sunscreen-image.json) renders the attachment right-aligned, same 28px corners, above the text bubble.
Assistant turns. Assistant prose is left-aligned plain text with no bubble. Markdown renders: bold/italic, ## headings, bullet and ordered lists (06-short-howto.json), --- horizontal rules and the [[cite:…]] pill chip (02-sleep-apnea-long.json), an [[icon:💡]] inline prefix, and title as bold heading text (03-poem-iphone-mac.json, 05-seatgeek-gpt-chip.json). feedback: false hides the thumbs chips.
Empty state. When messages is empty (04-empty-typing.json) the centered OpenAI spiral hero renders instead of a conversation.
Composer. Default is the full-pill (999px) bar with +, input, mic, and a round 44×44 send button. With composer.text set (04-empty-typing.json) the send button is solid black and a caret follows the typed text. With streaming: true (02, 03) the send button shows the black stop square. With composer.chip (05-seatgeek-gpt-chip.json) the Apps-SDK GPT chip renders inside the composer above the input row.
Failure Modes
Symptom
Likely cause
What to do
Thread file not found: … then exit 1
--thread path is wrong or the cwd is not the skill dir
Pass an absolute path or cd into the skill directory first (render.js resolves the path against cwd)
Error: playwright not installed
npm install never ran in the skill dir
Run npm install, then npx playwright install chromium (see Setup)
Run hangs ~15s then page.goto … timeout
screenshot.js is waiting on networkidle for an asset that never loads (e.g. a remote/http image src)
Use a local relative src that resolves under the thread.json dir, or an inline SVG asset like examples/assets/sunscreen-sky.svg
Image attachment shows the broken-image glyph
user-image.src is relative but the file is missing on disk
render.js rewrites a relative src to file://<threadDir>/<src>; make sure the asset actually exists there
Send button stays gray after setting text
composer.text is "" (empty string) rather than a real value
Set composer.text to a non-empty string, or omit it to keep the placeholder state
PNG comes out the wrong size
.stage width/height in templates/chat.css or DEVICE_SCALE/VIEWPORT in screenshot.js was changed
Keep .stage at 750 × 1624 and deviceScaleFactor at 3 so output stays 2250 × 4872
Header right icons overlap the model tag on a long title
header.title is too wide for the centered slot
Shorten the title, drop the model tag, or switch header.style to "title-only"
[[cite:…]] or [[icon:…]] prints literally instead of rendering
The marker is mistyped (missing double brackets / colon)
Match the exact [[cite:Source +1]] / [[icon:💡]] form parsed by formatInline in generate.js
Markdown --- shows as text, not an <hr>
The --- line is not on its own line between blank lines
Put a blank line above and below the ---, as in 02-sleep-apnea-long.json
1---2name: create-chatgpt-mockup3description: Render pixel-accurate ChatGPT mobile (iOS) screen mockups in light mode from a thread JSON. Supports user text bubbles, user image attachments, assistant markdown prose, citation chips, the OpenAI spiral logo, the Apps-SDK GPT chip in the composer, and three header styles (model-tag, plain title, "Get Plus"). Fixed 9:16 viewport. Outputs HTML + PNG.4---56# create-chatgpt-mockup
78## Purpose
910Generate single-frame ChatGPT mobile screens that read as real iOS captures — accurate header chrome, user/assistant message styling, composer pill, status bar, and the OpenAI spiral mark. Designed for use inside ad creatives, social posts, video storyboards, and product mockups.
1112If you need an **animated, video-recorded chat reveal** (bubbles popping over time, SFX, music bed, end card), use the molecule `skills/molecules/create-chatgpt-video-ad`. This atom only produces a single still frame.
1314## Inputs
1516Compose a `thread.json` matching the schema below, then pass it to `render.js`.
1718### Thread JSON schema
1920```json
21{
22 "statusBar": {
23 "time": "9:41",
24 "dnd": false
25 },
26 "header": {
27 "style": "model-tag",
28 "title": "ChatGPT",
29 "model": "5.1",
30 "rightIcons": ["personPlus", "dottedCircle"]
31 },
32 "messages": [
33 { "type": "user-image", "src": "assets/photo.jpg", "aspect": "square" },
34 { "type": "user-text", "text": "Make this face sunscreen better" },
35 {
36 "type": "assistant",
37 "title": "Optional H1 with spiral logo",
38 "feedback": true,
39 "text": "Markdown-ish prose. **bold**, *italic*, ## H2, * bullet, 1. ordered, [[cite:Source +1]], [[icon:💡]] inline-icon."
40 }
41 ],
42 "composer": {
43 "placeholder": "Ask anything",
44 "text": "Typed text to show with caret",
45 "cursor": true,
46 "streaming": false,
47 "chip": { "name": "SeatGeek" }
48 }
49}
50```
5152### Field reference
5354**`statusBar`**
55- `time` — string shown top-left (e.g. `"9:41"`, `"08:18"`).
56- `dnd` — when `true`, renders a small moon glyph next to the time (matches iOS Focus-on indicator).
5758**`header.style`** — one of:
59- `"model-tag"` — `≡ ChatGPT 5.1 ▾ ●●` — used when a specific model is active.
60- `"title-only"` — `≡ ChatGPT › ●●` — the plain in-chat header.
61- `"plain-title"` — `≡ Get Plus ✦ ●●` — for the upsell/empty-state title.
6263**`header.rightIcons`** — array of icon keys, rendered left-to-right on the right side. Available: `"personPlus"`, `"dottedCircle"`, `"edit"`, `"more"`. Default is `["personPlus", "dottedCircle"]` for `model-tag`, `["edit"]` for `title-only`.
6465**`messages[]`** — one of:
66- `{ "type": "user-image", "src": <path>, "aspect": "square"? }` — relative `src` paths resolve against the thread.json directory.
67- `{ "type": "user-text", "text": <string> }` — right-aligned soft-gray bubble.
68- `{ "type": "assistant", "text": <markdown>, "title"?: <string>, "feedback"?: <bool> }` — left-aligned plain prose. If `title` is set, an OpenAI spiral logo + bold title is rendered above the body. `feedback: false` hides the thumbs-up/down chips.
6970**Assistant `text` markdown** — supported syntax:
71- `**bold**`, `*italic*`
72- `# H1`, `## H2`, `### H3`
73- `* item` / `- item` bullet lists
74- `1. item` ordered lists
75- `---` horizontal rule
76- `[[cite:Some Source +1]]` — pill citation chip
77- `[[icon:💡]]` — inline emoji prefix (kept un-escaped)
78- Single `\n` → soft line break (`<br>`); double `\n\n` → new paragraph
7980**`composer`**
81- `placeholder` — shown when `text` is empty (default `"Ask anything"`).
82- `text` — typed-in text. When present, the send button activates (black bg).
83- `cursor` — whether to render the blinking caret after `text` (default `true`).
84- `streaming` — when `true`, replaces the send arrow with a stop square (matches "response generating" state).
85- `chip` — optional GPT chip in the composer. `chip.name` is the GPT name. Sets `placeholder` to `"Ask ChatGPT"` if you don't override it.
8687## CLI
8889```bash
90node render.js --thread examples/01-sunscreen-image.json
91node render.js --thread my-thread.json --output ./exports --name nightly
92```
9394### Flags
9596| Flag | Effect |
97|---|---|
98| `--thread <path>` | (required) JSON file matching the schema above |
99| `--output <dir>` | parent dir for the dated output folder; default `./chatgpt-mockup-exports/` |
100| `--name <slug>` | override the output folder slug |
101102## Output
103104```
105<output>/<YYYY-MM-DD>-<slug>/
106 index.html # standalone HTML (CSS inlined)
107 screenshot.png # 2250 × 4002 PNG (DPR 3, 9:16)
108 thread.json # copy of the input for reproducibility
109```
110111## Workflow
1121131. Receive a brief or existing thread JSON.
1142. If only a brief is given, compose `thread.json` matching the schema above.
1153. Run `node render.js --thread <path>`.
1164. Open the PNG. If anything looks off (header alignment, bubble width, missing icon), edit the JSON or `templates/chat.css` and re-render.
117118## Setup (one-time)
119120```bash
121cd create-chatgpt-mockup
122npm install
123npx playwright install chromium
124```
125126## Quality checks
127128- [ ] Renders in **light mode only** (white background, dark text). Dark-mode rendering is out of scope for v1.
129- [ ] Status bar time and right cluster (signal / wifi / battery) sit at the very top, no clipping under the iOS status-bar safe area.
130- [ ] Header chrome icons (hamburger / personPlus / dottedCircle / edit / more) all visually weigh the same as a real ChatGPT iOS screenshot at the same display size — **eyeball, don't pixel-match**. Sub-reading-weight icons read as broken even when their pixel dimensions "match the spec."
131- [ ] Header model tag (`5.1`, `5.2`) sits inline with `ChatGPT` and its chevron — no wrapping.
132- [ ] User bubble is **right-aligned**, soft-gray (`#F4F4F4`), `border-radius: 28px`, max-width 78% of stage.
133- [ ] User image attachments use the same `28px` corner radius and sit right-aligned above the text bubble.
134- [ ] Assistant body is plain prose (no bubble), left-aligned, full-width within the gutter.
135- [ ] **Composer corner radius is 999px** (full pill), `1px` border `#E5E5E5`, send button is `44 × 44` round.
136- [ ] When `composer.text` is set: send button is **solid black** with a white up-arrow. Caret sits right after the last character of typed text.
137- [ ] When `composer.streaming: true`: send button shows a **black square stop icon**.
138- [ ] When a citation marker `[[cite:Foo +1]]` is in the assistant text, it renders as a pill chip inline at the end of the sentence.
139140## Failure modes
141142| Symptom | Cause | Fix |
143|---|---|---|
144| `Thread file not found` | Wrong path or running from a different cwd | Use absolute paths or run from the skill directory |
145| Image attachment shows broken-image glyph | `src` is relative but the renderer couldn't resolve it | render.js resolves relative `src` against the thread.json's directory — make sure the asset exists at that path |
146| Send button stays light gray after typing | `composer.text` is empty string `""` instead of unset | Either omit `text` entirely or set it to a real string |
147| Caret renders as a stray `\|` separated from the text | Caret span is being treated as a flex child | The CSS rule for `.composer .input` uses inline content flow — do not switch it to `display: flex` |
148| Streamed list-item words render with no spaces ("Howconversationswork") | `.word { display: inline-block }` collapses inter-span text-node whitespace inside `<div>` parents (works fine inside `<p>`) | Use `display: inline` for `.word` — modern browsers honor `filter: blur()` + opacity transitions on inline elements too |
149| Bullets visible while list-item text is still hidden during streaming | CSS `list-style` markers are browser-painted; the word-stream wrapper only sees text nodes | The atom auto-converts `<ul><li>X</li></ul>` to `<div class="md-list"><div class="md-li">• X</div></div>` when `stream: true` — keep that pass; don't regress it. Any new markdown that emits browser-painted glyphs (`::marker`, `quotes`, `counter-content`) needs the same text-conversion treatment. |
150| Markdown `---` doesn't render an `<hr>` | The `---` line isn't on a line by itself, or the surrounding text didn't end its paragraph | Surround `---` with blank lines on both sides |
151| Stanzas of a poem joined into single paragraphs | Lines separated by single `\n` get joined as a single paragraph but with `<br>` between them — this **is** the intended behavior. If you want stanza breaks, use `\n\n` between stanzas. |
152| Header right cluster overlaps the model tag on long titles | Title text is too wide for the centered slot | Shorten `header.title`, drop the `model` tag, or switch `header.style` to `"title-only"` |
153154## Files
155156| File | Purpose |
157|---|---|
158| `render.js` | CLI entry — parses flags, generates HTML, takes screenshot, writes outputs |
159| `generate.js` | Thread JSON → standalone HTML page (markdown-ish parser inside) |
160| `screenshot.js` | HTML → PNG via Playwright (chromium headless) at fixed 750×1334 / DPR 3 |
161| `templates/chat.css` | All visual styling |
162| `templates/icons.js` | Inline SVG icons (hamburger, chevrons, spiral logo, mic, plus, send arrow, stop square, signal/wifi/battery, thumbs up/down) |
163| `examples/*.json` | 6 reference threads exercising every variant |
164| `examples/assets/` | Placeholder image assets used by image examples |
165| `tests/run-all.sh` | Render every example into `tests/output/` for visual review |
166167## Known limitations
168169- **Light mode only.** Dark mode would require a separate theme block and is out of scope for v1.
170- **No carousel / horizontal-scroll cards.** The Apps-SDK SeatGeek reference shows a horizontal card carousel inside the assistant response; not modeled here.
171- **No live keyboard rendering.** The composer is the bottom pill only — no on-screen keyboard chrome, suggestion bar, or QWERTY grid. (If you need the keyboard, scale the screen down 50% and overlay a separate keyboard image.)
172- **Markdown is intentionally minimal** — no tables, no fenced code blocks with syntax highlighting, no nested lists. Add to `generate.js#renderMarkdown` if a brief requires more.
173- **Emoji rendering** depends on system fonts. Chromium on macOS ships Apple Color Emoji; on Linux ships Noto Color Emoji.
174175## Quality Checks
176177Run `bash tests/run-all.sh` and open the PNGs in `tests/output/`, then confirm:
178179- **Every example renders without error.** All six `examples/*.json` produce an `index.html` + `screenshot.png` + `thread.json` triplet; `render.js` exits `0`.
180- **Dimensions are stable.** Each `screenshot.png` is **2250 × 4872** (the `.stage` is `750 × 1624` and `screenshot.js` captures at `deviceScaleFactor: 3`). A wrong size means the stage box or DPR changed.
181- **Light mode, white background.** Background is white, text near-black. This atom does not render dark mode.
182- **Status bar.** Time sits top-left; the signal / wifi / battery cluster sits top-right, unclipped. When `statusBar.dnd: true` (see `02-sleep-apnea-long.json`, `05-seatgeek-gpt-chip.json`) a moon glyph renders next to the time.
183- **Header style matches the JSON.** `model-tag` shows `ChatGPT 5.1 ▾` (`01-sunscreen-image.json`, `06-short-howto.json`); `title-only` shows the plain title + right chevron (`02`, `03`, `05`); `plain-title` shows the upsell title with no chevron (`04-empty-typing.json`). The center cluster never wraps and the right icons never overlap it.
184- **User turns.** User text is a right-aligned soft-gray bubble (`#F4F4F4`, `28px` radius). A `user-image` (see `01-sunscreen-image.json`) renders the attachment right-aligned, same `28px` corners, above the text bubble.
185- **Assistant turns.** Assistant prose is left-aligned plain text with no bubble. Markdown renders: bold/italic, `##` headings, bullet and ordered lists (`06-short-howto.json`), `---` horizontal rules and the `[[cite:…]]` pill chip (`02-sleep-apnea-long.json`), an `[[icon:💡]]` inline prefix, and `title` as bold heading text (`03-poem-iphone-mac.json`, `05-seatgeek-gpt-chip.json`). `feedback: false` hides the thumbs chips.
186- **Empty state.** When `messages` is empty (`04-empty-typing.json`) the centered OpenAI spiral hero renders instead of a conversation.
187- **Composer.** Default is the full-pill (`999px`) bar with `+`, input, mic, and a round `44×44` send button. With `composer.text` set (`04-empty-typing.json`) the send button is solid black and a caret follows the typed text. With `streaming: true` (`02`, `03`) the send button shows the black stop square. With `composer.chip` (`05-seatgeek-gpt-chip.json`) the Apps-SDK GPT chip renders inside the composer above the input row.
188189## Failure Modes
190191| Symptom | Likely cause | What to do |
192|---|---|---|
193| `Thread file not found: …` then exit `1` | `--thread` path is wrong or the cwd is not the skill dir | Pass an absolute path or `cd` into the skill directory first (`render.js` resolves the path against cwd) |
194| `Error: playwright not installed` | `npm install` never ran in the skill dir | Run `npm install`, then `npx playwright install chromium` (see Setup) |
195| Run hangs ~15s then `page.goto … timeout` | `screenshot.js` is waiting on `networkidle` for an asset that never loads (e.g. a remote/`http` image `src`) | Use a local relative `src` that resolves under the thread.json dir, or an inline SVG asset like `examples/assets/sunscreen-sky.svg` |
196| Image attachment shows the broken-image glyph | `user-image.src` is relative but the file is missing on disk | `render.js` rewrites a relative `src` to `file://<threadDir>/<src>`; make sure the asset actually exists there |
197| Send button stays gray after setting text | `composer.text` is `""` (empty string) rather than a real value | Set `composer.text` to a non-empty string, or omit it to keep the placeholder state |
198| PNG comes out the wrong size | `.stage` width/height in `templates/chat.css` or `DEVICE_SCALE`/`VIEWPORT` in `screenshot.js` was changed | Keep `.stage` at `750 × 1624` and `deviceScaleFactor` at `3` so output stays `2250 × 4872` |
199| Header right icons overlap the model tag on a long title | `header.title` is too wide for the centered slot | Shorten the title, drop the `model` tag, or switch `header.style` to `"title-only"` |
200| `[[cite:…]]` or `[[icon:…]]` prints literally instead of rendering | The marker is mistyped (missing double brackets / colon) | Match the exact `[[cite:Source +1]]` / `[[icon:💡]]` form parsed by `formatInline` in `generate.js` |
201| Markdown `---` shows as text, not an `<hr>` | The `---` line is not on its own line between blank lines | Put a blank line above and below the `---`, as in `02-sleep-apnea-long.json` |
Run npx skillmds add majiayu000/create-chatgpt-mockup in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.
Render pixel-accurate ChatGPT mobile (iOS) screen mockups in light mode from a thread JSON. Supports user text bubbles, user image attachments, assistant markdown prose, citation chips, the OpenAI spiral logo, the Apps-SDK GPT chip in the composer, and three header styles (model-tag, plain title, "Get Plus"). Fixed 9:16 viewport. Outputs HTML + PNG. It is listed under Design & Media on SkillMD.
This skill has not completed SkillMD's automated safety review yet. Capability flags: docs only. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.
This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.
Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.
majiayu000 (@majiayu000) published this skill. Their other Agent Skills are listed on their SkillMD profile.