Connect · Open interfaces

Bring GoodCase into your workflow.

Agent discoveryOne entry point for agents: llms.txtOpen →

01 · Agent Skill

Install the GoodCase Skill in your AI

The Skill reads the same public API as the site. Install it through the same cross-agent installer used by the LearnPrompt Skill Workshop:

npx skills add LearnPrompt/goodcase-lite --skill goodcase --global --copy --yes --full-depth

Once installed, ask your AI “What good cases are new?” and it can query GoodCase directly.

02 · RSS

Subscribe to new cases with RSS

Add this URL to any RSS reader. Each item includes a localized summary and prompt preview, newest first.

https://goodcase.ai/en/feed.xml

After subscribing

Search “goodcase” in your reader to find this feed again — or hand the URL straight to your AI and have it read it for you every morning.

03 · REST API

Call the open REST API

GET /api/public/cases · Case list

curl -s "https://goodcase.ai/api/public/cases?category=video&q=umesh&take=5&locale=en"

category: Filter by image / video / web / copy / hardware. Other values return 400.
q: Case-insensitive keyword match across title, summary, and creator.
take: Return 1–50 items; defaults to 20 and clamps out-of-range values.
locale: Content language: zh-CN or en. Defaults to zh-CN.
form: Secondary form filter: text-to-video / ref-to-video / code-generated-video / web-ui / motion-3d / game-3d / agent-tool. Other values return 400.
sample: Deterministic random sample size, 1–50, clamped out-of-range. Overrides take when present.
seed: Sampling seed, used with sample. Same seed + params reproduces the same batch; omitted seeds are random and echoed back in the response.

Response fields (per item) · 15 fields→

slug: Unique case identifier used in detail URLs.
title / summary: Localized title and summary.
category: image / video / web / copy / hardware
creator / source / sourceUrl: Creator name, source platform, and original URL.
sourceLikeCount / sourceCommentCount / sourceShareCount / sourceSaveCount: Public interaction snapshot; unavailable platform fields are null.
sourcePublishedAt / sourceMetricsCapturedAt: Source publish time and metrics capture time.
sourceHeatScore / sourceInteractionCount: Cross-platform normalized heat and raw interaction count; null without a snapshot.
currentHeatScore: sourceHeatScore decayed by publish time (7-day half-life, 15% floor after 60 days); null when sourceHeatScore is null.
promptPreview: Prompt preview; the full prompt is in the detail endpoint.
mediaUrl / posterUrl: Absolute media and poster URLs.
stabilityScore: Current stability reference.
favoriteScore: Legacy compatibility field; not used by Source Heat Ranking.
recommendedModels / costBand: Recommended models and low / medium / high cost band.
evidenceLevel / tags: L0 / L1 / L2 evidence level and tags.
url: Localized case detail URL.

GET /api/public/cases/{slug} · Full case

curl -s "https://goodcase.ai/api/public/cases/real-case-01-umesh-ai?locale=en"

Adds promptFull, editorNote, labNote, spreadScore, sourceHeatNote, and other deep-reading fields to the list response.

GET /api/public/cases/{slug}/retests · Retest evidence

Returns this case's on-site per-model retest records — stabilityScore, evidenceLevel, per-model latest/history, and the full record list. Use it to compare a model you just tested against the site's own baseline.

All interfaces are anonymous and require no API key. Server responses cache for five minutes and allow direct browser fetches through CORS.

§ Model Eval · Model eval

Model eval: test any new model with the same set.

The day a new model ships, an agent with the goodcase Skill installed can sample cases by category and run them, then compare against the site's own retest baseline.

What it does under the hood→

01Sample: GET /api/public/cases?category=web&form=web-ui&sample=5&seed=20260906-yourmodel. Defaults to exclude=motion-3d — background video / 3D tests the asset, not the model. Same seed, same result, so recording the seed makes it reproducible.
02Pull the baseline: GET /api/public/cases/{slug} for promptFull, GET /api/public/cases/{slug}/retests for the site's own per-model retest records.
03Run and compare: run promptFull on the target model, record reproduced / degraded / failed per case, and output a comparison table plus a goodcase-model-eval-v1 JSON list — the seed + slug list is what lets a later on-site retest line up against this same batch.

First install the goodcase skill: npx skills add LearnPrompt/goodcase-lite --skill goodcase --global --copy --yes --full-depth
Then evaluate <model> with goodcase: sample 5 cases with category=web, form=web-ui (seed=20260906-<model>), pull promptFull and the on-site /retests baseline for each case, run every prompt on <model>, record reproduced / degraded / failed per case, then give me a comparison table plus a goodcase-model-eval-v1 JSON. This is a self-run test, not an official goodcase retest.

A self-run result is not an official goodcase retest; finalScore (judge score) and stabilityScore (human review tier) are not on the same scale.

No access to codex / claude / grok? goodcase skill ships the same judging rubric it uses on-site — the reproduced / degraded / failed criteria — so any model can score against it. Report judge="self" plus the model you used; self-run results are shown separately from the official three-judge results and never averaged together.

§ Method · Method

From scattered hits to reusable methods.

A · Original / scatteredInternet

@servasyy_ai · X / 𝕏

A real coding/UI case that links to the open-source 3DCellForge project, turning image input into an interactive 3D workflow.

Prompt excerpt
https://github.com/huangserva/3DCellForge...

#AIcase #prompt #model #workflow #creator

→

B · Structured case pageGoodCase.ai

Image to 3D

AI Coding · X / 𝕏

Signal

SourceX / 𝕏 · @servasyy_ai

Prompt41 chars · standardized preview

Models testedClaude Sonnet · GPT-5 · Gemini Pro

Creator@servasyy_ai · pattern owner

Signal

Stable

Cost

MID

Retest record

Image to 3D · Beta evidence

§ Open Source · Open source & more

Open source, and the extended family.

GoodCase Lite (open-source site)

The open-source site of GoodCase.ai — clone it and run the same UI, with real sample cases built in.

GitHub ↗

Awesome Seedance (verified prompt library)

Verified Seedance 2.5/2.0 prompt library — every prompt human-checked against its original post, with author, source, date, and heat score.

GitHub ↗

§ About · Product boundary

One product. One case source of truth.

Case

The public unit: work, creator, method, and original source.

Creator

A creator view aggregated from accepted cases.

Lab

Retest records and stability evidence around a case.

Skill

A method package distilled after repeated success across cases.

§ Feedback · Feedback

Found a mistake? Leave it here.

Content corrections, page issues, and product suggestions enter a separate human queue and never masquerade as case submissions.

Join the GoodCase Feishu group

New cases, retest results and submission status land in the group first. Authors and operators are both there.

Join the Feishu group →

Loading…

Install the GoodCase Skill in your AI

The Skill reads the same public API as the site. Install it through the same cross-agent installer used by the LearnPrompt Skill Workshop:

npx skills add LearnPrompt/goodcase-lite --skill goodcase --global --copy --yes --full-depth

Once installed, ask your AI “What good cases are new?” and it can query GoodCase directly.

Call the open REST API

GET /api/public/cases · Case list

curl -s "https://goodcase.ai/api/public/cases?category=video&q=umesh&take=5&locale=en"

category: Filter by image / video / web / copy / hardware. Other values return 400.
q: Case-insensitive keyword match across title, summary, and creator.
take: Return 1–50 items; defaults to 20 and clamps out-of-range values.
locale: Content language: zh-CN or en. Defaults to zh-CN.
form: Secondary form filter: text-to-video / ref-to-video / code-generated-video / web-ui / motion-3d / game-3d / agent-tool. Other values return 400.
sample: Deterministic random sample size, 1–50, clamped out-of-range. Overrides take when present.
seed: Sampling seed, used with sample. Same seed + params reproduces the same batch; omitted seeds are random and echoed back in the response.

Response fields (per item) · 15 fields→

slug: Unique case identifier used in detail URLs.
title / summary: Localized title and summary.
category: image / video / web / copy / hardware
creator / source / sourceUrl: Creator name, source platform, and original URL.
sourceLikeCount / sourceCommentCount / sourceShareCount / sourceSaveCount: Public interaction snapshot; unavailable platform fields are null.
sourcePublishedAt / sourceMetricsCapturedAt: Source publish time and metrics capture time.
sourceHeatScore / sourceInteractionCount: Cross-platform normalized heat and raw interaction count; null without a snapshot.
currentHeatScore: sourceHeatScore decayed by publish time (7-day half-life, 15% floor after 60 days); null when sourceHeatScore is null.
promptPreview: Prompt preview; the full prompt is in the detail endpoint.
mediaUrl / posterUrl: Absolute media and poster URLs.
stabilityScore: Current stability reference.
favoriteScore: Legacy compatibility field; not used by Source Heat Ranking.
recommendedModels / costBand: Recommended models and low / medium / high cost band.
evidenceLevel / tags: L0 / L1 / L2 evidence level and tags.
url: Localized case detail URL.

GET /api/public/cases/{slug} · Full case

curl -s "https://goodcase.ai/api/public/cases/real-case-01-umesh-ai?locale=en"

Adds promptFull, editorNote, labNote, spreadScore, sourceHeatNote, and other deep-reading fields to the list response.

GET /api/public/cases/{slug}/retests · Retest evidence

First install the goodcase skill: npx skills add LearnPrompt/goodcase-lite --skill goodcase --global --copy --yes --full-depth Then evaluate <model> with goodcase: sample 5 cases with category=web, form=web-ui (seed=20260906-<model>), pull promptFull and the on-site /retests baseline for each case, run every prompt on <model>, record reproduced / degraded / failed per case, then give me a comparison table plus a goodcase-model-eval-v1 JSON. This is a self-run test, not an official goodcase retest.