image-generation

AI 图片生成：原生 tool call（Codex/Antigravity）或浏览器自动化（Gemini/ChatGPT）。 Use when: 需要 AI 生成概念图、UI 参考、像素画素材、完整 PPT 页面、复杂架构图、信息图或视觉 mock。 Not for: 已有图片的展示（用 media_gallery rich block）、硬要求可编辑/native text 的 PPT/图表（用 PPT/HTML 管线）。 Output: 生成图片自动发布，或作为完整视觉 mock / 图像素材进入后续交付。

Quality

88%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

77%Weight 40%Scale 1-3

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

Highly actionable and well-sequenced content with concrete selectors, paths, and feedback-gated workflows, but it is monolithic with no bundle references and repeats its core imagegen-first guidance across several sections. Tightening the redundancy and splitting the reference tables would raise the weaker dimensions.

Suggestions

Consolidate the repeated 'imagegen-first / do not write SVG/HTML first' guidance into a single canonical section (e.g., the SVG 复发熔断闸) and have other sections reference it, removing ~4 restatements to improve conciseness.

Move the DOM selector tables and file-naming tables into a reference bundle file (e.g., references/browser_selectors.md) and link to it one level deep, so the body stays an overview and progressive disclosure improves.

Add an explicit validation step in the browser paths (e.g., confirm the generated image element is present before clicking download) so every workflow has a verification checkpoint.

Dimension	Reasoning	Score
Conciseness	Operational and mostly free of basic-concept padding, but the 'default to full-page imagegen, do not write SVG/HTML first' message is restated across at least five sections (能力校准, SVG 复发熔断闸, Full-page raster first, 什么时候才写 SVG, Common Mistakes, 注意事项), so it could be tightened. Not a 3 because of this redundancy; not a 1 because it does not explain concepts Claude already knows and most tokens earn their place.	2 / 3
Actionability	Provides fully concrete, copy-paste-ready guidance: exact tool names ('image_gen', 'generate_image'), exact paths ('~/.codex/generated_images/<sessionId>/'), precise DOM selectors ('.ql-editor[contenteditable="true"]', '#prompt-textarea'), button labels, filename formats, numbered browser steps, and a shell example. Not a 2 because the examples are executable rather than pseudocode.	3 / 3
Workflow Clarity	Sequences are explicit with validation checkpoints and feedback loops — '视觉评估' → prompt iterate → '失败才降级' after 1-2 attempts, plus the 'SVG override reason' gate and the Common Mistakes table acting as a checklist. Not a 2 because error-recovery checkpoints are explicit rather than implicit.	3 / 3
Progressive Disclosure	No bundle files exist and the ~245-line body is monolithic; reference-style content (DOM selector tables, file-naming tables, platform matrix) is kept inline with no one-level-deep references. Not a 1 because sections are well-organized rather than a poorly-organized wall; not a 3 because nothing is split into referenced files for a skill well over 50 lines.	2 / 3
	Total	10 / 12 Passed

Description

100%Weight 40%Scale 1-3

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong description: it states concrete capabilities, provides explicit 'Use when:' triggers, and disambiguates with a 'Not for:' clause. It is concise yet covers methods, platforms, and deliverables without fluff.

Dimension	Reasoning	Score
Specificity	Lists concrete actions and deliverables — '原生 tool call（Codex/Antigravity）或浏览器自动化（Gemini/ChatGPT）' plus '概念图、UI 参考、像素画素材、完整 PPT 页面、复杂架构图、信息图或视觉 mock' — naming both methods and a comprehensive set of outputs, matching the score-3 anchor for multiple specific concrete actions. It is not a 2 because coverage is comprehensive rather than partial.	3 / 3
Completeness	Explicitly answers both what (AI image generation via native tool call or browser automation) and when via a literal 'Use when:' clause, plus a 'Not for:' anti-trigger and 'Output:' line. Not a 2 because the when is explicit, not merely implied.	3 / 3
Trigger Term Quality	The 'Use when:' clause surfaces natural terms users would say — '概念图、UI 参考、像素画素材、完整 PPT 页面、复杂架构图、信息图、视觉 mock' — alongside platform names (Codex/Antigravity/Gemini/ChatGPT), giving good coverage of natural trigger vocabulary. Not a 2 because common variations are well represented rather than missing.	3 / 3
Distinctiveness Conflict Risk	The 'Not for:' clause distinguishes it from existing-image display ('用 media_gallery rich block') and editable-deck pipelines ('用 PPT/HTML 管线'), carving a clear niche unlikely to trigger the wrong skill. Not a 2 because the anti-triggers remove overlap with adjacent skills.	3 / 3
	Total	12 / 12 Passed

Validation

100%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: zts212653/clowder-ai
Path: cat-cafe-skills/image-generation/SKILL.md
Commit: 80782c5

Reviewed: 11 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.