Aetheria-Labs1/storefront-skills / skills/ab-test
ab-test
VerifiedAnalyze an editable Lexsis landing-page URL, plan one controlled change, build verified challenger pages, and create or evaluate a draft A/B test. Use for URL-first storefront experimentation, not ordinary page editing.
references/workflows/section-asset-workflow.md
View on GitHub# Section asset workflow
This is the single `/plan-assets` execution procedure for sourcing,
missing-slot handling, and visual fit. Type files provide jobs, tags, crops and budgets; policy
lives in `references/assets/asset-sourcing-sequence.md`,
`references/assets/generation-policy.md`, `references/assets/slot-spec.md`,
`references/assets/video-rules.md` and the relevant proof ledger.
## 1. Acquisition and missing-slot handling
1. Read the type's section job and budget. Decide whether imagery is needed;
a native disclosure or factual table does not need a decorative image.
2. Inventory real catalogue media, the merchant's selections, and existing
library assets before planning new media. Assign each candidate its actual
job, product/variant, dimensions and rights evidence.
3. Follow the source eligibility order in
`references/assets/asset-sourcing-sequence.md`. Use the type's search tags
and queries. A missing tag is not proof that the library is empty.
4. Run the fit review in section 2 before binding a candidate. Retain rejection
reasons so the same unsuitable image is not proposed repeatedly.
5. For a remaining slot, consult `references/assets/generation-policy.md`.
That file owns ALLOW / ASK / NEVER, credit approval, identity protection,
prompts and generation records. Do not infer permission from a page type.
6. Tell the merchant what is missing: section, job, aspect, count and why it
matters. Offer **upload**, **generate** only where that policy permits it,
or **skip/merge** with an explanation of what the page loses. Group gaps
into one useful message, not a question for each section.
7. A required production slot remains unresolved until a verified source is
bound. Do not silently remove a required section or substitute a
placeholder. A section adjustment returns to `/plan-page`.
8. Skip or merge a section only on the merchant's decision. A generation ban
is not permission to invent media, substitute irrelevant stock, or hide
the missing job behind a decorative band.
### Import, upload and selection
- `lexsis_asset_library.search` returns read-only search evidence: the terms
and filters used, matching assets, result count, and pagination state. Do not
wait after search.
- `lexsis_asset_select.select` opens the interactive picker when the merchant
should choose among fit-reviewed candidates. Pass the relevant search terms
and filters, wait for the user's selection message, and map its order to the
named slots. It may be called directly, although searching first is the
normal agent workflow.
- `lexsis_asset_import.import` persists an available URL, image base64 with
MIME type, or conversation attachments. Supply exactly one source.
- `lexsis_asset_upload.upload` opens the local-file UI. Scope it to the
selected workspace/theme and wait for the user's uploaded-asset message.
Without inline UI, ask for a URL or conversation attachment and import it.
Opening a panel is not an upload. The exact argument contract is in
`references/lexsis-mcp-contract.md`.
### Search efficiently
Use the current search schema. Tags such as `hero`, `lifestyle`,
`product-shot`, `social-proof` and `logo` are conventions, not a closed enum.
Try semantic queries for missing jobs and filename lookup for a supplied
filename. Where supported, OCR search helps identify baked-in overlays and
similarity search around a verified seed helps maintain a coherent set.
Result geometry screens candidates before the visual review; metadata never
replaces inspection. Keep the selected workspace and theme binding explicit.
## 2. View and fit review
Nothing is used sight unseen. Do not accept the closest existing asset merely
because it is available. Open each candidate with `lexsis_assets.view`;
if the host cannot display it, inspect the returned permanent URL with the
available image viewer. Judge the asset **in its intended section**:
| Check | Required decision |
|---|---|
| Identity and job | Correct product/variant; the image actually demonstrates the assigned job |
| Crop | Desktop and mobile crops preserve the subject and required detail |
| Text placement | Copy has an appropriate quiet region, or moves outside the image; A7 still governs contrast |
| Set consistency | View adjacent/grid candidates together; they should read as one shoot rather than unrelated finds |
| Palette | The image works with the plan's visual direction without falsifying product appearance |
| Resolution | Detail survives at the actual mobile and desktop rendered sizes |
| Source integrity | Rights are recorded; no misleading overlay, watermark or competitor branding; genuine product labels remain readable where the job requires them |
A near-uniform preview can be a failed preview. Inspect the original URL
before rejecting the asset or spending credits on a replacement. Apply this
same review after generation; payment is not evidence of suitability. A
rejected generation returns to the policy's bounded repair/fallback route.
## 3. Asset budget and record
The type's budget records supplied jobs, missing jobs and type-specific
alternatives. It does not redefine source permissions. Assign each slot a
source decision, rights basis, final asset or Shopify media id, and status.
Use the binding and production tables defined by `/plan-assets`; generation
records follow the generation-policy owner. Mark rejected or unresolved jobs
accurately.
Example merchant message: "The kit has a packaging image but no image of
all included items. Supply one overhead kit photo, or choose to omit the
optional unboxing section; the included-items job is still pending."
## 4. Write to the resolved job
After the media decision, execute `references/workflows/copy-workflow.md`.
Do not invent benefit imagery or pad the copy because a requested image is
missing. A deliberately text-only section follows its type contract.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |