Home / All Skills / Content & Media Creation / canghe-cover-image
English (EN)简体中文 (ZH)日本語 (JA)
Content & Media Creation · freestylefly/canghe-skills

canghe-cover-image

Generates article cover images with 5 dimensions (type, palette, rendering, text, mood) combining 9 color palettes and 6 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
Author freestyleflyGitHub freestylefly/canghe-skills ↗License MIT（仓库根 LICENSE）Commit dd0bf35595
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 having value enumerations and default values, ultimately assembling a structured prompt.

skills/canghe-cover-image/SKILL.md
Generate elegant cover images for articles with 5-dimensional customization.
What is happening here: The Five Dimensions table provides values and defaults for each dimension (Text defaults to title-only, Mood defaults to balanced, Font defaults to clean, Aspect defaults to 16:9); Font is the 6th actual dimension (not counted in the summary description).

Provides an automatic selection mapping table of "content signal -> value" when dimensions are missing.

skills/canghe-cover-image/SKILL.md
Auto-selection rules: [references/auto-selection.md](references/auto-selection.md)
What is happening here: Maps content type, palette, rendering style, text density, mood, and font one-by-one based on keyword hits; automatic selection serves as the foundation for --quick mode.

Confirmation is a soft gate: skipped only with --quick or when all dimensions are specified; aspect ratio is still asked unless explicitly provided via --aspect.

skills/canghe-cover-image/SKILL.md
| `--quick` or `quick_mode: true` | 6 dimensions | Aspect ratio (unless `--aspect`) |
What is happening here: Step 2 uses a condition table to decide which dimensions to skip; aspect ratio is the only item that may still be prompted even under --quick.

Persists the prompt to disk as prompts/cover.md (including frontmatter references) prior to generation.

skills/canghe-cover-image/SKILL.md
**CRITICAL - References in Frontmatter**:
What is happening here: Step 3 requires reference images saved to refs/ to be documented in frontmatter references, omitted if style is only verbally extracted; test -f is used for self-check before writing.

Routes reference images by purpose: direct pass-through via --ref, while style/palette features are extracted and appended to the prompt.

skills/canghe-cover-image/SKILL.md
- `direct` usage → pass via `--ref` (use ref-capable backend)
What is happening here: Step 4 item 3 handles images separately based on their frontmatter references purpose, preventing all reference images from being treated as direct pass-through inputs.

Output directory and artifact structure are fixed, defaulting to an independent folder at cover-image/{topic-slug}/.

skills/canghe-cover-image/SKILL.md
- `independent` (default): `cover-image/{topic-slug}/`
What is happening here: The directory contains source, refs/, prompts/cover.md, and cover.png; same-dir / imgs-subdir preferences are also optionally available.

Enforces strict composition and text rules: 40-60% negative space, simplified silhouettes for characters, and exact user/original text for the title.

skills/canghe-cover-image/SKILL.md
- **Whitespace**: 40-60% breathing room
What is happening here: The Composition Principles section also specifies centering or left-aligning the visual anchor, using simplified silhouettes for characters, and strictly using exact original text for the title (never invent).

Backs up existing assets before regeneration, requiring the prompt file to be updated prior to re-generating.

skills/canghe-cover-image/SKILL.md
| **Regenerate** | Backup → Update prompt file FIRST → Regenerate |
What is happening here: The Image Modification table establishes the fixed procedure of "backup -> update prompt -> regenerate" to ensure traceability of changes.

2Core Capabilities

01Six-dimensional cover customization + preset shortcuts (specify all at once via --style)
02Automatic selection based on content signals (9 palettes / 6 renderings / 6 types / text density / mood / font)
03Structured prompt template (sectioned + frontmatter references)
04Reference image handling: direct pass-through via --ref, feature extraction into prompt for style/palette
05Pre/post-generation validation and backup (test -f reference images, cover.png backup, single retry on failure)
06Multi-aspect ratio support (16:9 / 2.35:1 / 4:3 / 3:2 / 1:1 / 3:4)
07Text and composition discipline (exact original title, silhouettes only for characters, 40-60% negative space)
08Preference persistence (EXTEND.md: watermark / preferred_* / default_aspect / quick_mode / language)

3External Dependencies

KindDependency / Endpoint
api图像生成 skill（Step 4 在宿主可用 skill 中挑选；多个时问用户偏好，SKILL.md 未点名具体后端）

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 read/write and artifact output: Agent writes to disk driven by SKILL.md instructions, saving source-{slug}, refs/ under cover-image/{topic-slug}/
  • SKILL.md: Declares invocation of host CLI interpreters or external execution environments
Audit Notes: (a) The skill's stated purpose is to generate cover images from article content: Step 4 instructs the agent to 'Check image generation skills' and hand off the prompt file, output path, and aspect ratio to the selected image generation skill/backend. Executing this will send the article title/summary/keywords and reference images externally to a third-party image model, consuming account credits. (b) Local disk writes are explicit: source, refs/, prompts/cover.md, cover.png under cover-image/{topic-slug}/, as well as EXTEND.md. (c) The skill's own directory contains zero scripts, zero native network endpoints, and zero credential reading; network egress and credentials reside entirely within the delegated image generation backend, so it is not classified as red. While this document does not name a specific backend (only stating 'Check image generation skills'), the designed workflow inherently sends materials to external generation services. Following precedents such as baoyu-cover-image / ai-image-generation where assets are dispatched to a generation backend, this is classified as orange.

5Second-Pass Verification

  • [ok] 自身目录无任何脚本或可执行文件 — 文件清单只有 .md；扫描“执行外部程序/写盘/网络请求”均为 0 命中，脚本执行面为空。
  • [discrepancy] 外部生成后端名称 — SKILL.md 仅写 'Check image generation skills; if multiple, ask preference'，未点名具体后端；出网事实由 skill 的声明用途（生成封面）推出，档位仍为 orange。
  • [ok] EXTEND.md 查找只探测存在性 — $HOME 命中仅是 test -f 的存在性判断，未读取文件内容之外的凭证或环境变量。
  • [ok] 维度取值集合一致 — SKILL.md Five Dimensions 表与尺寸/palette 目录文件数对齐；prompt 模板与确认流程均存在。
  • [ok] 参考图分流与自检 — Step 3/4 要求 frontmatter 列 references、写入前 test -f、direct 走 --ref、style/palette 抽取特征，规则闭环。
  • [ok] meta 元数据 — commit dd0bf355…、path skills/canghe-cover-image、license MIT 与 bundle §0 核定值一致。

6Verdict & Scenarios

  • Layout-level customization across six dimensions (including font/mood/text density), offering far more control than simple "generate from a topic" approaches.
  • Dual-path confirmation process: default multi-turn confirmation alongside --quick / quick_mode skipping, balancing reliability and speed.
  • Explicit reference image processing: direct pass-through via --ref, style-based feature extraction into prompts, with enforced file existence checks.
  • Strict text and composition constraints, preventing the model from hallucinating titles or cluttering elements.
    • Best for: Writers or content teams who need to generate stylized article/blog covers with controllable dimensions like palette, rendering, and text volume, alongside reproducible prompts.
    Not for: Strictly offline environments or scenarios where external data egress is prohibited; batch generation of multiple in-article illustrations (use canghe-article-illustrator instead); zero-confirmation, strictly one-click generation scenarios.
    Install Agent-Ready
    ① Verified Mirror 2026-10-10
    Option A · Direct DownloadDownload canghe-cover-image.tar.gz
    sha256: 92d75c156ff10b20…
    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 dd0bf35595.