agentleFS
wangge-aiwangge-ai/wangge-skills/skills/ecom-main-image-diagnosis/SKILL.md

ecom-main-image-diagnosis

Skill4 starsChanged 39 days agoMITView on GitHub

Use when diagnosing or improving Chinese ecommerce main images, search-result cards, product-page screenshots, ad creatives, competitor image sets, designer briefs, AI image prompts, CTR problems, click hooks, visual hierarchy, platform fit, or main-image A/B test plans for Taobao, Tmall, JD, Pinduoduo, Douyin, Xiaohongshu, 1688, or similar marketplaces.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If visual material is unavailable, ask for an image, screenshot, or rendered page before making visual claims.

    line 32
  2. Distinctive:

    Run the click-gap map:

    line 39
  3. Distinctive:

    Test next: 2-3 variants with one clear hypothesis each.

    line 48
  4. Distinctive:

    Prefer "test this hypothesis" over "this will improve CTR".

    line 56
  5. Distinctive:

    Never reuse source-branded report headers, examples, names, or footers from external skill packages.

    line 59
  6. Distinctive:

    Read only the references needed for the user's requested output.

    line 70

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Ecom Main Image Diagnosis

Core Stance

Diagnose the user's buying decision at the image-card level, not whether the image is "nice". The useful question is:

Will the right buyer notice it, understand it, trust it, and have a sharper reason to click than nearby alternatives?

This is a clean-room local skill. Do not include external author names, toolbox names, contact details, private version labels, generated-by footers, or source authorization text in reports, examples, README files, or share packages.

First Triage

Classify the input before diagnosing:

InputUse It ForCaveat
Single main imageQuick click diagnosis and minimal edit planMark product facts as inferred
Search-result screenshotShelf contrast, price-band pressure, visual crowdingCompare only visible competitors
Product-page screenshotMain image plus title, price, SKU, trust signalsDo not infer backend data
1-5 image setImage sequence task planningJudge each image's job, not decoration
Competitor imagesDifferentiation map and visual crowdingDo not copy competitor assets or text
Product linkPublic field and image acquisitionIf blocked, ask for screenshot or saved page
CTR/CVR/exposure dataTest design and priority settingDo not promise future uplift

If visual material is unavailable, ask for an image, screenshot, or rendered page before making visual claims.

Diagnosis Workflow

  1. State the evidence scope: what was visible, what was inferred, and what is missing.
  2. Identify the platform scene: search shelf, recommendation feed, product page, ad placement, Xiaohongshu cover, or mixed use.
  3. Build a compact strategy card: product, buyer, price band, decision pressure, strongest click promise, strongest proof.
  4. Run the click-gap map:
    • Recognition gap: can buyers tell what it is in one glance?
    • Benefit gap: is the buyer outcome concrete?
    • Trust gap: does it reduce risk before click?
    • Difference gap: does it avoid looking like the rest of the shelf?
    • Focus gap: does the product beat background, labels, and props?
    • Platform gap: does the style match where it will be shown?
  5. Produce a minimum viable change plan:
    • Fix now: 1-3 edits that remove the biggest click blockers.
    • Test next: 2-3 variants with one clear hypothesis each.
    • Rebuild later: deeper reshoot, copy, proof, or sequence changes.
  6. Offer optional outputs only when useful: designer brief, AI image prompt, 1-5 image sequence, A/B test plan, or share-ready case summary.

Output Rules

  • Be concrete enough for a designer or operator to act: placement, scale, crop, copy length, color role, proof asset, and safe area.
  • Separate facts, visual observations, inferences, and suggestions.
  • Prefer "test this hypothesis" over "this will improve CTR".
  • For formal reports, use neutral metadata only: material source, platform scene, generated time, assumptions, and limits.
  • For public articles or share packages, lead with the result and usage value. Keep commands, validation logs, blocked attempts, and regeneration notes in README or handoff docs.
  • Never reuse source-branded report headers, examples, names, or footers from external skill packages.

Reference Routing

  • For input routing and evidence labels, read references/input-router.md.
  • For the click-gap framework and scoring, read references/diagnosis-framework.md.
  • For platform-specific notes, read references/platform-playbook-cn.md.
  • For report, quick diagnosis, designer brief, prompt, and test formats, read references/output-contract.md.
  • For AI prompt and designer handoff details, read references/prompt-and-brief.md.
  • For public sharing, article packaging, and source-neutral release checks, read references/share-guidance.md.

Read only the references needed for the user's requested output.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.