Design & Motion · JimLiu/baoyu-skills

baoyu-cover-image

Name: baoyu-cover-image
Author: JimLiu/baoyu-skills

Generates article cover images with 5 dimensions (type, palette, rendering, text, mood) combining 11 color palettes and 7 rendering styles. Supports cinematic (2.35:1), widescreen (16:9), and square (1:1) aspects. Use when user asks to "generate cover image", "create article cover", or "make cover".

L4 High-Privilege · Graded auth & full auditV1 Commit-PinnedExecute CommandWrite FileDual-pass Audit

Control levels describe skill capabilities and suggested control measures; they do not represent quality or safety certifications, nor do they dictate usability. Usage decisions rest entirely with users and administrators. Different agent hosts apply varying constraints; do not assume hosts provide automatic protection. Third-party skill copyrights belong to their original authors. Use at your own risk.

1How It Works & Mechanism

Deconstructs 'what the cover looks like' into six confirmable dimensions (Type/Palette/Rendering/Text/Mood/Font) plus aspect ratio, each with an enumerated set of values and defaults, ultimately assembling them into a structured prompt.

skills/baoyu-cover-image/SKILL.md

Generate elegant cover images for articles with 5-dimensional customization.

The Five Dimensions table in SKILL.md provides values and defaults for each dimension (Text defaults to title-only, Mood to balanced, Font to clean, Aspect to 16:9); Font is effectively the sixth dimension (omitted from the initial count in the description).

Features an automatic selection table for 'content signals → values' when dimensions are omitted: keyword matching maps content type, color palette, rendering style, text density, mood, and font one by one.

skills/baoyu-cover-image/references/auto-selection.md

When a dimension is omitted, select based on content signals.

For example, 'Education, tutorial, knowledge, onboarding, concept explainer → macaron' and 'Poster, movie, album, concert, silhouette, mondo, limited-edition → screen-print'; fallback defaults are centered around Text/Mood/Font.

User confirmation acts as a hard gate: host interactive tools (AskUserQuestion or equivalent) must be used to present options as clickable/selectable questions rather than simply dumping a raw text table.

skills/baoyu-cover-image/SKILL.md

**MUST use `AskUserQuestion` tool** to present options as interactive selection — NOT plain text tables.

Batched into a maximum of 4 questions; skipped only when --quick / quick_mode: true is set or all six dimensions plus --aspect are fully specified. The aspect ratio must always be asked unless explicitly supplied via --aspect.

Persists the prompt to prompts/cover.md before generation (structured into content context / visual design / text elements / mood / font / composition sections), including a references list in the frontmatter.

skills/baoyu-cover-image/SKILL.md

Save to `prompts/cover.md`.

Requires pre-write validation that reference image files exist (test -f); the body restricts color values to 'rendering guidelines' and strictly prohibits rendering hex codes or color names as visible text on the image.

Treats reference images as 'high-priority visual inputs' with an explicit warning that passing --ref alone is insufficient: what to replicate must be concurrently described in the prompt text.

skills/baoyu-cover-image/references/workflow/prompt-template.md

**Passing `--ref` alone is NOT enough.**

For models that do not support --ref (specifically naming Jimeng and Seedream 3.0), it falls back to writing per-character text descriptions in refs/ref-NN-{slug}.md and embedding them into the prompt with MUST/REQUIRED prefixes.

Backend resolution follows the repository's unified hierarchy: requested name > EXTEND.md preference > automatic selection (Codex imagegen > codex-imagegen > Cursor GenerateImage > other native > only installed backend > ask user).

skills/baoyu-cover-image/SKILL.md

**Codex (`imagegen`)** — first, inspect your available-skills / tool inventory.

Further demonstrates that the skill delegates to a single generation backend; during generation, the prompt file, output path, and aspect ratio are passed to that backend, with an automatic retry on failure.

Enforces strict composition and text discipline: 40-60% negative space, simplified silhouettes for characters, exact verbatim titles from the user/source without hallucination; text errors must be resolved by modifying the prompt and regenerating rather than inpainting.

skills/baoyu-cover-image/SKILL.md

**Title**: Use exact title from user/source; never invent

The same section explicitly states 'Characters: Simplified silhouettes; NO realistic humans'; the correction path specifies 'regenerate from a corrected prompt, switch to a lower-text or no-title variant'.

Standardizes the output directory and artifact structure: by default saves independently to cover-image/{topic-slug}/ containing source, refs/, prompts/cover.md, and cover.png, backing up the existing image before regenerating.

skills/baoyu-cover-image/SKILL.md

- `independent` (default): `cover-image/{topic-slug}/`

Step 4, Item 1 mandates 'Backup existing cover.png if regenerating'; slugs use 2-4 kebab-case words, appending a timestamp in case of conflict.

2Core Capabilities

01Six-dimensional cover customization + preset shortcuts (--style configures all at once)

02Automatic dimension selection based on content signals (11 palettes / 7 renderings / 6 types / text density / mood / font)

03Structured prompt templating (Content Context / Visual Design / Text Elements / Composition sections + frontmatter references)

04Reference image handling: direct passing for models supporting --ref, translated into per-character text descriptions for unsupported models

05Pre/post-generation validation and backups (test -f for reference images, backup of cover.png, automatic retry on failure)

06Multi-aspect ratio support (16:9 / 2.35:1 / 4:3 / 3:2 / 1:1 / 3:4), always prompted unless explicitly specified

07Placeholder and color value discipline (color values are rendering guides only, forbidden from appearing as visible text)

08Preference persistence (EXTEND.md: watermark / preferred_* / default_aspect / quick_mode / language)

3External Dependencies

Kind	Dependency / Endpoint
api	Codex 原生 imagegen skill（Skill 工具调用）
cli	codex CLI / baoyu-image-gen（codex-imagegen 分支）
api	Cursor 原生 GenerateImage / Hermes image_generate
network	codex-imagegen wrapper（自述认证走用户 Codex 订阅）

4Recommended Governance & Evidence

Suggested Controls for Safe Operation

For Agent Runtime

Graded/per-use authorization, sandbox recommended

For Administrators

Approval workflow, rollback support, full audit

Evidence & Triggers (File Paths & Excerpts)

本地工作目录: File I/O and artifact output: Written to disk by the agent per SKILL.md instructions: source-{slug}, refs/ (references) under cover-image/{topic-slug}/
SKILL.md: Declares invocation of host CLI interpreters or external execution environments

Audit Notes: (a) SKILL.md explicitly names external generation backends to invoke at runtime: '**Codex (`imagegen`)** — first, inspect your available-skills / tool inventory.' and 'route through `baoyu-image-gen --provider codex-cli` (preferred)' (same rules apply for Cursor `GenerateImage` and Hermes `image_generate`). Following SKILL.md sends article titles, summaries, keywords, and reference images out to third-party image models, consuming user quota. (b) Local disk writes are explicit: prompts/cover.md, refs/, cover.png, and EXTEND.md. (c) The skill directory itself contains zero scripts, zero built-in network endpoints, and zero credential reads; network egress and credentials reside entirely within the invoked backends, so it is not marked red. In line with precedents (e.g., ai-image-generation, nano-banana-2), skills sending material to generation backends are uniformly rated orange.

5Second-Pass Verification

[ok] 自身无脚本/无端点/无凭证 — 目录内全是 markdown；URL 扫描只命中 github.com/JimLiu/baoyu-skills。
[ok] SKILL.md 点名生成后端（orange 依据） — Codex imagegen / codex CLI via baoyu-image-gen / Cursor GenerateImage / Hermes image_generate 均在 SKILL.md 正文，且写成“一次选定、生成前必须完成”。
[ok] 维度数量说法核对（描述写 5 dimensions） — SKILL.md 正文 Actually 列 Type/Palette/Rendering/Text/Mood 五维于 Five Dimensions 表，Font 单列在枚举与 Options 中——描述与正文不冲突，侦查按正文六项记录。
[ok] 文字/色值纪律 — prompt-template.md 的 Color constraint 与 SKILL.md 的 Title/Characters/Composition Principles 三处对得上；位图涂改禁令与 article-illustrator 同款。
[ok] 参考图不支持 --ref 时的降级路径 — SKILL.md Step 1 与 reference-images.md 一致：写 refs/ref-NN-{slug}.md 逐人物描述，并以 MUST/REQUIRED 嵌入 prompt。
[ok] meta 元数据 — GitHub API：MIT、25926 stars、pushed_at 2026-09-10T15:13:43Z；本地 HEAD 与 pin 1567581c26ec… 一致。

6Verdict & Scenarios

High controllability: explicit confirmation across six dimensions and aspect ratio ensures reproducible styles and unified batch outputs.

Content-aware defaults: automatically selects palettes, rendering styles, and fonts based on keywords when parameters are omitted, offering a smooth cold-start experience.

Actionable reference images: templates require reference elements to be formulated as MUST/REQUIRED directives, preventing models from ignoring --ref.

Reliable text: titles are copied verbatim from the source and color values are prohibited from appearing on-canvas, mitigating garbled text and hallucinated titles.

Dual-speed workflow: --quick / quick_mode persistently skips confirmation, while the interactive default prevents unexpected cover designs.

Best for: Ideal for bloggers, WeChat Official Account creators, and technical writers who need a consistent visual standard for cover images, producing them in batches by topic while preserving color and typographic coherence; also well-suited for teams standardizing cover design decisions into structured workflows.

Not for: Not suitable for covers requiring pixel-perfect, precise typography (which should be handled by dedicated graphic design tools); not suitable for environments strictly prohibiting data egress; not suitable for non-interactive hosts that cannot accept default values.

Install Agent-Ready

① Verified Mirror 2026-09-15

Option A · Direct DownloadDownload baoyu-cover-image.tar.gz

sha256: e68671656994141a…

Option B · Copy Guide for Agent

install.json

Agents automatically download and verify mirrors from this JSON guide.

② Upstream GitHub

GitHub Repo ↗

Pinned to commit 1567581c26.