gooseworks-ai/goose-skills / drafts/marketing/make-product-photos
make-product-photos
VerifiedUse when the marketer says: studio photos of my product without a shoot, product photos for my store page, packshots on white, lifestyle or on-model shots, photos for the new colourway, fresh product imagery for ads, photos for our launch page, our packaging changed — redo the photos, I have no photographer. Delivers store-ready product photos generated from the brand's real product photos: a shot list built from the page slots (hero, alt angle, scale, detail, in use, lifestyle, on-model), only real photos of this product as references, matched to the brand's photography, every shot checked against the real product for shape, parts, logo, text, colour, finish, scale and honest representation, with a keep / maybe / reject verdict and a page order. The human picks the keepers. Never uploads to a store, never shows a size, part, feature or result the product lacks. Not for ads with text, infographics or screenshots — make-static-ads; not for marketplace listing sets (not covered yet).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 15.4 KB |
| atoms/fidelity-check/ATOM.md | markdown | 2.4 KB |
| atoms/shot-list/ATOM.md | markdown | 1.6 KB |
| atoms/source-audit/ATOM.md | markdown | 4.2 KB |
| references/fidelity-rubric.md | markdown | 6.1 KB |
| references/fit-and-honesty.md | markdown | 5.8 KB |
| references/shot-list-and-surfaces.md | markdown | 7.8 KB |
| skill.meta.json | config | 3.0 KB |
| tests/test-by-hand.md | markdown | 6.5 KB |
references/fidelity-rubric.md
View on GitHub# Fidelity rubric — checking outputs against the real product
Load at Step 7. The engine's judge checks silhouette, materials, logo and colourway; this check is worth something only if it is **independent and more specific**. The engine's `passed` is necessary, not sufficient; `flagged` is shown to the human with its reason.
**What each output gives you** (`photos_read` `generation_id`): `status`, `final_image_url`, `thumbnail_url`, `quality_score`, `final_defect_code` / `final_defect_message`, `suggested_control_change`, `credits_charged`, `media_asset_id`, `attempts[]`; the generation carries `ratio`, `quality`, `overridden_controls`.
## 1. Checks, in order (stop at the first blocker)
| # | Check | How to look | Severity |
|---|---|---|---|
| F12 | **Honesty** | The `fit-and-honesty.md` §2 list: implied features, contents, results, size, food styling, invented marks | Blocker regardless of beauty |
| F2 | **Parts** | Count caps, lids, loops, straps, buttons, zips, pockets, laces, handles, pump heads — nothing added, missing or duplicated | Blocker |
| F3 | **Logo** | Right mark, place and size; crisp; not mirrored or re-lettered | Blocker |
| F4 | **Label and printed text** | Read every legible string by eye against the real label photo; OCR is a hint (staging read the same label as "51.4 ML" and "61.4 ML"). Text-shaped blur is OK only where the real text is also tiny | Blocker |
| F1 | **Silhouette and proportions** | Outline and key ratios (lid:body, strap length, collar, heel height) vs the same-angle real photo | Blocker |
| F5 | **Colour** | Same colourway under the scene's light; hero compared under neutral light; no bleed between variants | Blocker for hero/alts; major for lifestyle |
| F7 | **Scale** | Size vs hands, bodies, furniture, pockets plausible for the real dimensions (ask for them if the record has none) | Blocker for scale / in-use |
| F9 | **Hands, bodies, faces** | Finger count, joints, nails, grip pressure, skin–product boundary, eyes, teeth, jewellery merging | Blocker if visible |
| F8 | **Physics** | One light direction; shadows away from it; contact (no floating unless styled); reflections match the scene | Major |
| F10 | **Background** | Stray or duplicate products, ghost text, another brand's logo, warped lines, banding. A host device the product fits (a watch for a band) is allowed when shown accurately | Major; blocker if an unrelated brand appears |
| F6 | **Material and finish** | Matte vs gloss, metal type, weave and knit, transparency | Major (honesty if it implies a better material) |
| F11 | **AI look** | Plastic sheen, waxy skin, bokeh cutting through objects (airbrushed skin beside a skincare product is F12 `misleading`) | Taste; major if the house style is natural |
| F13 | **House style** | Background, light side, shadow, angle, crop and padding, grade, retouch, casting vs the style card; grid test against 3–4 existing heroes | Major for hero/alts; minor for mood |
## 2. Named reject reasons
Use exactly one of: `wrong_shape` · `extra_or_missing_part` · `logo_wrong` · `text_garbled` · `colour_off` · `finish_wrong` · `scale_wrong` · `physics` · `hands_or_body` · `background_artefact` · `ai_look` · `misleading` · `off_style` — plus a specific note ("text_garbled: back label at small sizes"). Use them in the verdict you show the human. Archive with no `reason` — the engine would turn it into a brand-wide "dont". A durable rule the human states goes through the shared save-learning atom.
## 3. When you can see images
Output and the same-angle real photo side by side; zoom on the logo, the label and any hands. Record which checks you did by eye.
## 4. When you cannot
Follow `../../../../skills/marketing/honest-review/SKILL.md` §1–3.
- Run `gooseworks:media_analyze` (`action: "read"`) on the output's `media_asset_id` **and** on the same-angle reference photo — both free.
- Compare the two reads **field by field** and report each mismatch specifically:
| Field | From the reads | A mismatch looks like |
|---|---|---|
| Part and item count | lids, loops, straps, caps, items in a set named in the description | reference "white cap", output "two caps" |
| Label and printed text | `visible_text` | "DIVE DEEP" vs "DIVE DEEEP"; a mark the label does not carry |
| Colour words | the description's colour terms | "green bottle" vs "mint bottle" |
| Product identity and size | product type, size text (e.g. "51.4 ML") | a different product, or the mini instead of the full size |
| Extra things in frame | other products, other logos, people, hands | a second brand's logo; hands with no attestation |
- OCR misreads (staging read one label as both "51.4 ML" and "61.4 ML"): a text mismatch is "please check", not proof.
- Label the photo **"not checked by eye"** — never present the engine's pass as your own check — and give the human a three-item checklist aimed at what the reads could not settle (finish, scale, physics, the exact logo shape).
## 5. Calibrate first
Run the rubric once on one of the brand's real **product-only** photos (not an infographic or a lifestyle composite). If you would reject it, you are too strict (or reading the wrong reference).
## 6. Keeper vs reject — the water-bottle example
*32 oz matte powder-coat insulated bottle, colourway "Sage", laser-etched logo, chug lid with a carry loop, does not sweat.*
- **Hero keeper:** outline matches the real photo (lid ≈ 1/6 of body height); Sage reads as the same muted green under neutral light; broad soft highlight, no mirror reflections; logo correct, crisp, right size and place; one loop on the correct side; one contact shadow consistent with a left key; clean house background.
- **Hero reject:** a generic taller cap (`wrong_shape`); chrome highlights on a matte bottle (`finish_wrong`); warped logo letters (`logo_wrong`); mint or olive instead of Sage (`colour_off`); two loops (`extra_or_missing_part`); shadow falling toward the light (`physics`).
- **Scale keeper:** bottle-to-hand ratio plausible for a 27 cm bottle; five fingers; a real grip.
- **In-use reject:** condensation beads on the insulated bottle (`misleading` — an honesty fail, not taste); a second brand's logo on the backpack (`background_artefact`).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |