Skip to content
Back to search

image-studio

SkillWorks with Claude CodeGitHubIncomplete

Turn a business website link into a ready-to-use set of marketing images that show what the business actually sells. Reads the site, captures real proof of the work (client sites, product UI, demo elements) and frames it in browser and phone mockups, harvests its best existing photos, fixes the near-misses with Codex edits, generates whatever is missing with Codex (free, ChatGPT quota), rejects anything that looks like AI slop or could be any business, and crops each image to the size it will be used at. Delivers exactly N images into a new folder in Downloads. Use for "/image-studio", "images for this site", "marketing images for <link>", "give me N images for this business", "social images from my website", "find the best images on this site", "make images for this brand", "edit these images for Instagram", or any request for a set of brand or marketing images (not video: that is creator-studio).

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/adarsh350--image-studio/badge.svg)](/listings/adarsh350--image-studio)

HTML

<a href="/listings/adarsh350--image-studio"><img src="/api/listings/adarsh350--image-studio/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

SKILL.md · Manifest · 241 lines
  1. ---
  2. name: image-studio
  3. description: Turn a business website link into a ready-to-use set of marketing images that show what the business actually sells. Reads the site, captures real proof of the work (client sites, product UI, demo elements) and frames it in browser and phone mockups, harvests its best existing photos, fixes the near-misses with Codex edits, generates whatever is missing with Codex (free, ChatGPT quota), rejects anything that looks like AI slop or could be any business, and crops each image to the size it will be used at. Delivers exactly N images into a new folder in Downloads. Use for "/image-studio", "images for this site", "marketing images for <link>", "give me N images for this business", "social images from my website", "find the best images on this site", "make images for this brand", "edit these images for Instagram", or any request for a set of brand or marketing images (not video: that is creator-studio).
  4. argument-hint: "<url> [count] [--use social|web|ads|mixed] [--proof <folder>] [--existing-only]"
  5. ---
  6. # Image Studio
  7. Link in, N finished images out. `<skill-dir>` is this skill's base directory. First time on a
  8. machine: `npm install` in `<skill-dir>` (Playwright; the Chromium it needs is shared with
  9. creator-studio, else `npx playwright install chromium`). Needs Node 20+, Python 3.10+ with Pillow,
  10. and the Codex CLI or the OpenAI VS Code extension signed in to ChatGPT.
  11. Read these before writing the brief. The quality of the set is decided there.
  12. - `<skill-dir>/references/playbook.md`: what to make. Placements and their specs, worked prompts
  13. for 12 business types, and the default mix for "N images".
  14. - `<skill-dir>/references/craft.md`: how to make and judge each image. The brief template, the
  15. prompt recipe, style anchors and the three-pass reject gate.
  16. - `<skill-dir>/references/industries/INDEX.md`: maps any business to one of 52 industry sections
  17. (benchmark brands, house style, slop tells, shot concepts, real-only shots), with a fallback for
  18. a business no section fits.
  19. **The bar:** every delivered image can go straight into a paid ad or onto a client's Instagram
  20. today. If it would look cheap next to the photos in `references/anchors/`, it isn't delivered.
  21. And the set sells: at least 60% of it shows the business's real work or product, so a stranger
  22. scrolling past could say what the business does (craft.md section 2).
  23. ## 1. Intake
  24. Need three things: the **link**, the **count**, and the **use**. If the invocation gave all three,
  25. don't ask. Otherwise ask once with AskUserQuestion, only for what is missing:
  26. | Question | Options |
  27. |---|---|
  28. | How many images? | 6 · 10 · 20 (Other for any number) |
  29. | What are they for? | Mixed social + web (Recommended) · Social posts · Website · Ads |
  30. `--proof <folder>` (or screenshots the user attaches or points to): real screenshots of the work,
  31. such as a product dashboard, a client site, a report. Read every file in it. Rank them by how
  32. clearly each shows one offer, how sharp and large it is, and whether it holds personal data
  33. (skip those unless the owner cleared them). Then use the best one or two per offer as `proof`
  34. slots (craft.md section 2). Copy any proof file the set uses into `$W/proof/`. Never generate a
  35. stand-in for proof the user said they will supply: leave the slot open and say so.
  36. Defaults when the user says "just do it": 10 images, mixed. Whose brand: a client site gets no
  37. Iyara context; Iyara's own site loads `iyara-labs` for the brief. If the count is over 20, say it
  38. draws on the ChatGPT allowance (about 35s and one message per generated image, plus retries) and
  39. continue.
  40. **Sizes by use** (each slot gets one):
  41. | Use | Size | Ratio |
  42. |---|---|---|
  43. | Feed post (Instagram, LinkedIn, Facebook) | 1080x1350 | 4:5 |
  44. | Square post / profile tile | 1080x1080 | 1:1 |
  45. | Story, Reel cover, TikTok | 1080x1920 | 9:16 |
  46. | Website hero, blog header | 1920x1080 | 16:9 |
  47. | Link preview (OG), ad landscape | 1200x630 | 1.91:1 |
  48. Mixed set: use the default set in playbook.md section 4. It includes the Google Business Profile
  49. slots, which take **real photos only** (Google's policy bars stock imagery, and a generated
  50. photo of premises misrepresents them). If the site has no real photo for a GBP slot, the slot
  51. becomes a feed post, and the reply lists the GBP shots the owner should take.
  52. ## 2. Read the site
  53. ```bash
  54. W="$LOCALAPPDATA/image-studio/runs/<slug>-$(date +%Y-%m-%d-%H%M%S)" # ~/.local/share/image-studio/runs on macOS/Linux
  55. node <skill-dir>/scripts/harvest.mjs "<url>" "$W/site" --pages 6
  56. ```
  57. Writes `site.json` (pages, headings, nav, meta, fonts, palette, every image with size and role),
  58. `home.png` and `images/`. Then:
  59. - Read `home.png` and the 6-10 largest `photo` images (contact sheet first:
  60. `python -I <skill-dir>/scripts/finish.py sheet "$W/site/sheet.jpg" <files>`).
  61. - Write `$W/brief.md` per craft.md section 1, including **Offers and proof**.
  62. - Sites behind bot protection or empty shells: say so, retry once with `--pages 1`, else ask for
  63. screenshots or a folder of images and continue from the brief.
  64. **Capture the proof.** For each offer in the brief, screenshot the real thing:
  65. the client sites it links to (`site.json` `pages[].outbound`, plus the work or case-study pages),
  66. and the UI or demo elements on its own pages (`pages[].ui`: a selector and a hint). View
  67. `home.png` first and pick only what is plainly the work, not navigation or a footer.
  68. ```bash
  69. node <skill-dir>/scripts/capture.mjs --batch "$W/capture.json"
  70. # capture.json: [{"url": "https://client-site.com", "out": "<W>/cap/client-desktop.png"},
  71. # {"url": "https://client-site.com", "out": "<W>/cap/client-mobile.png", "device": "mobile"},
  72. # {"url": "https://client-site.com", "out": "<W>/cap/client-full.png", "full": true},
  73. # {"url": "<url>", "out": "<W>/cap/demo.png", "selector": "<pages[].ui selector>"}]
  74. ```
  75. desktop is 1440x900 at 2x and mobile 390x844 at 3x; cookie banners and chat widgets are hidden.
  76. Read every capture: a half-loaded page, a surviving banner or placeholder copy is recaptured
  77. (raise `--wait`) or dropped.
  78. Work shown only as lazy-loaded demo videos (a case-studies page with no outbound links) is still
  79. proof: scroll the page so the `<video>` sources load, download them, and pull frames with
  80. `ffmpeg -ss <t> -i v.mp4 -frames:v 1 f.png`. Screen recordings almost always carry the mouse
  81. pointer. Pick the frame where it sits over flat colour and paint it out with that colour
  82. (cleanup, not an edit). If it covers copy or numbers, use another frame.
  83. ## 3. Shot list
  84. Find the business in `references/industries/INDEX.md` and read its section (INDEX.md says how to
  85. use each part and what to do when nothing fits). Its House style goes into the brief and every
  86. prompt; its Never list joins the reject gate; its Real only shots are never `generate`.
  87. Write `$W/shots.md` per craft.md section 2: N rows of `#, use, size, shows, concept, source, ref`.
  88. Assign showcase slots first (at least 60% of N, every offer once when N allows), from the proof,
  89. the section's Showcase concepts and playbook.md section 3. Then fill mood slots (at most 40%) from
  90. the Shot concepts.
  91. Post it in chat as a compact table and **carry on without waiting** (the user asked for a link-in,
  92. files-out flow). `--existing-only` means no `generate` rows: if the site lacks N good photos,
  93. deliver fewer and say how many it had.
  94. Harvested images are good for `existing` only when all of these hold:
  95. - they are photos, not graphics;
  96. - they are sharp;
  97. - they have no baked-in text;
  98. - they are big enough that the final crop upscales 1.5x or less.
  99. Mine big site photos before generating anything. One photo with a long side of 3000px or more
  100. often holds two or three deliverables at its natural subject points: the wide scene, a person, a
  101. detail. Each crop is its own `existing` row with its own `--focus`. Logos are never a slot.
  102. ## 4. Make
  103. - **Proof, capture, existing:** straight to finishing (or into a mockup).
  104. - **Mockup:** frame captures on the brand ground at the slot's delivery size, all in one batch:
  105. ```bash
  106. node <skill-dir>/scripts/mockup.mjs --batch "$W/mockup.json"
  107. # mockup.json: [{"layout": "pair", "out": "<W>/raw/02.png", "size": "1080x1350",
  108. # "imgs": ["<W>/cap/client-desktop.png", "<W>/cap/client-mobile.png"],
  109. # "bg": "<brand ground hex>", "url": "client-site.com"},
  110. # {"layout": "split", "out": "<W>/raw/05.png", "size": "1920x1080",
  111. # "imgs": ["<W>/cap/before.png", "<W>/cap/after.png"], "labels": ["Before", "After"]}]
  112. ```
  113. Layouts: `browser` (1 desktop capture), `phone` (1 mobile), `pair` (desktop + mobile), `split`
  114. (before + after), `element` (1 UI element). Options: `bg`, `fg` (label colour), `ground` (a
  115. texture image), `chrome` light|dark, `url` (the real domain, the only text drawn besides the
  116. labels), `margin`. Output is 2x the size, for `finish.py fit` to bring down. A `WARN` (capture
  117. upscaled past 1.5x) means a bigger capture or a different layout.
  118. - **Screen:** generate the device scene with a flat green screen (craft.md "Screen scenes"), then
  119. `python -I <skill-dir>/scripts/finish.py screen <scene> <capture> "$W/raw/NN.png"`.
  120. - **Edit and generate:** write prompts per craft.md section 3 into `$W/jobs.json` and run all of
  121. them in one batch:
  122. ```bash
  123. python -I <skill-dir>/scripts/codex_image.py --batch "$W/jobs.json" --workers 3
  124. # jobs.json: [{"prompt": "...", "out": "<W>/raw/04a.png", "aspect": "4:5",
  125. # "style_refs": ["<skill-dir>/references/anchors/hands-craft-fresh-pasta.jpg"]},
  126. # {"prompt": "extend the scene left and right...", "out": "<W>/raw/07.png", "aspect": "16:9",
  127. # "refs": ["<W>/site/images/007.jpg"]}]
  128. ```
  129. The script lints every prompt first and spends no quota if any prompt fails. It then wraps each
  130. brief in Codex instructions: a labelled spec, the size for the aspect, the highest
  131. quality, and the fixed realism and exclusion clauses. Anchors (`style_refs`) come from
  132. `references/anchors/manifest.json`, by the shot's look first and the industry second (craft.md,
  133. "Style anchors").
  134. Make two candidates per `generate` slot so the gate has a choice: the same concept, one job with
  135. `style_refs` and one without. Anchors pull the grade toward the site; the plain run often keeps
  136. more fine detail. Tested 2026-11-19: each won one slot. If both candidates fail, rewrite the prompt. Ask Codex for the slot's own aspect: the built-in tool returns about 1.5MP at that aspect
  137. whatever size is requested (4:5 comes back 1122x1402), which is enough for every slot up to a
  138. 1920x1080 cover. The result's `size` field shows what came back, and `qa.py` catches anything too
  139. small.
  140. No Codex (`codex_image.py --check` fails): don't use a paid image API. Fall back to
  141. `--existing-only` and say so.
  142. ## 5. Gate
  143. Run craft.md section 4 on every candidate:
  144. 1. `qa.py` (machine pass);
  145. 2. the contact sheet;
  146. 3. full size, with `finish.py zoom` on every hand, face and possible lettering.
  147. Retry rejects as the gate says: rewrite once, try a new concept once, then fall back to the
  148. next-best harvested photo. After finishing, run `qa.py` once more on `final/` at each delivery
  149. size (it takes file paths, not a folder: `qa.py "$W"/final/0[1-5]-*.jpg --size 1080x1350`), and
  150. look at the final contact sheet as one set.
  151. ## 6. Finish and deliver
  152. ```bash
  153. python -I <skill-dir>/scripts/finish.py fit <source> "$W/final/01-feed-<concept>.jpg" --size 1080x1350 --focus 0.5,0.4
  154. ```
  155. Pick `--focus` from what you saw (the subject's position, as fractions of width and height). The
  156. script warns above a 1.5x upscale: then pick another source or regenerate, don't ship it soft.
  157. - **Deliver:** copy `final/` to `~/Downloads/<slug>-images-<YYYY-MM-DD>/` (add `-2`, `-3` if it
  158. exists). Exactly N images, named `NN-<use>-<concept>.jpg`. Nothing else in that folder.
  159. - **Chat reply:** the folder path; a table of `file, size, shows, source (site page URL / capture
  160. of <url> / supplied proof / Codex edit / Codex generated)`; any offer that had no proof, as a
  161. request for screenshots; anything the user should check (an edited real photo, a slot that fell back);
  162. the real-only and GBP shots the site lacked, as a short list for the owner to take; and the work
  163. dir for re-rolls.
  164. - **Rights:** images harvested from a site may be licensed stock that is cleared for that website
  165. only. Say so in one line when the site isn't the user's own or the set includes harvested photos.
  166. - **Re-rolls** ("redo 04", "warmer", "two more"): reuse the latest work dir for that slug, change
  167. only those slots, overwrite them in the same Downloads folder.
  168. - **Prune** at the start of each run: `find "$LOCALAPPDATA/image-studio/runs" -mindepth 1 -maxdepth 1 -type d -mtime +30 -exec rm -rf {} +`
  169. ## Hard rules
  170. - Never generate the client's logo, product as sold, team, premises, certificates, screenshots,
  171. product UI, dashboards, client work, AI answers or metrics. Those come from the site (captured),
  172. or from the user.
  173. - At least 60% of the set is showcase, and every showcase slot passes the stranger test.
  174. - Google Business Profile slots are real photos only.
  175. - Never put text, lettering or logos in a generated image. Real pixels (captures, supplied
  176. screenshots, the site's own copy) may carry their own text.
  177. - Never alter what a capture or supplied screenshot shows: crop, frame and composite it, never
  178. edit its numbers or copy.
  179. - Never ship an image that failed the gate to make up the count. Deliver fewer and say why.
  180. - Paid tools need a flag first: the Higgsfield MCP, any image API other than Codex, any paid
  181. upscaler.
  182. ## Related skills
  183. - `creator-studio`: WHEN the ask is a video. It can take this skill's finished images as stills.
  184. - `social` / `instagram-content` / `linkedin-content`: WHEN the images need captions and a
  185. posting plan. They consume the Downloads folder this skill writes.
  186. - `brand`: WHEN the site has no usable visual identity and one has to be defined first.
  187. ## Self-tests
  188. `node scripts/harvest.mjs --self-test` · `node scripts/capture.mjs --self-test` ·
  189. `node scripts/mockup.mjs --self-test` · `python -I scripts/finish.py --self-test` (fit, sheet,
  190. zoom and screen) ·
  191. `python -I scripts/qa.py --self-test` · `python -I scripts/codex_image.py --check` (lint and prompt tests
  192. plus the Codex version; no quota used). All at once: `npm run check`.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.