Design & Motion · JimLiu/baoyu-skills

baoyu-xhs-images

Name: baoyu-xhs-images
Author: JimLiu/baoyu-skills

Generates infographic image card series with 12 visual styles, 8 layouts, and 3 color palettes. Breaks content into 1-10 cartoon-style image cards optimized for social media engagement. Use when user mentions "小红书图片", "小红书种草", "小绿书", "微信图文", "微信贴图", "image cards", "图片卡片", baoyu-xhs-images, or wants social media infographic series.

L4 High-Privilege · Graded auth & full auditV1 Commit-PinnedExecute CommandNetwork AccessWrite 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

Breaks content into 1–10 image cards: analyzes audience/hook/selling points first, then determines each card's visual appearance using a 3D matrix of 'Style × Layout × Palette', and finally generates cards one by one.

skills/baoyu-xhs-images/SKILL.md

| **Style** | Visual aesthetics (lines, decorations, rendering) | 12 styles (see Styles below) |

Scale choices are straightforward: 12 styles × 8 layouts × 3 palettes, with approximately 25 presets collapsing popular combinations into a single parameter (--preset).

Image generation backends are not implemented within this repository: resolved via 'Current Request > EXTEND.md > Runtime Native Tools > Only Non-Native Backend > Ask User', explicitly naming Codex imagegen, codex-imagegen, Cursor GenerateImage, etc.

skills/baoyu-xhs-images/SKILL.md

If a skill named `imagegen` is listed, you are running inside Codex and MUST use it

Consequently, user content and prompts for each card are sent to the selected image backend; the skill itself contains no scripts or networking code.

The core technique for visual consistency is the 'Image 1 Anchor Chain': Card 1 (cover) is generated without a reference image, and every subsequent card passes Card 1 via --ref so characters, mascots, and palettes do not drift across images.

skills/baoyu-xhs-images/SKILL.md

**Visual consistency — image-1 anchor chain**: character / mascot / color rendering drifts between calls unless you anchor them.

Locking style via image-to-image is more stable than relying solely on sessionId (and the codex-imagegen wrapper does not support sessionId anyway).

Confirmation is divided into three paths: A (trust recommendation and generate immediately), B (customize step-by-step), and C (detailed mode: asks about selling points/audience first, then provides 3 distinct outline variants for user selection).

skills/baoyu-xhs-images/SKILL.md

**Path C — Detailed mode**: two sub-confirmations.

Step 2 is a hard gate, but allows explicit skipping via --yes or 'generate directly'; skipping requires first declaring assumed strategy, style, layout, palette, card count, and backend.

Prompt files strictly precede generation: before invoking any backend, the complete prompt for each image must first be written to prompts/NN-{type}-{slug}.md.

skills/baoyu-xhs-images/SKILL.md

**Prompt file requirement (hard)**: write each image's full, final prompt to a standalone file under `prompts/`

The prompt file is the sole basis for reproducibility and switching backends; modifying an image requires modifying the file before regenerating.

First-time use triggers a blocking preference setup: when EXTEND.md is missing in interactive mode, the initial setup must run before analyzing content.

skills/baoyu-xhs-images/references/config/first-time-setup.md

**⛔ BLOCKING OPERATION**: This setup MUST complete before ANY other workflow steps.

In non-interactive mode (--yes), setup is skipped entirely, built-in defaults are used, and EXTEND.md is not written—preventing hangs during automation.

Includes a style × layout compatibility matrix to display warnings when users select discouraged combinations; auto-selection matches content signal keywords against the first matching row, falling back to cute-share.

skills/baoyu-xhs-images/SKILL.md

Compatibility scores (✓✓ highly recommended, ✓ works well, ✗ avoid).

Combinations like study-notes×sparse or screen-print×dense are marked with ✗ in the matrix, turning design experience into a queryable lookup table.

Content breakdown follows a fixed skeleton: Cover (Hook) + Content Cards (each carrying a single value point) + Ending (CTA/Summary), reporting progress card-by-card and retrying only failed cards.

skills/baoyu-xhs-images/SKILL.md

| Cover (image 1) | Hook + visual impact | `sparse` |

Batch strategy shares lineage with slide-deck: Native backend batch > runtime parallelism (default 4) > serial; with an additional constraint of the 'image-1 anchor'.

2Core Capabilities

01Flexible image card generation combining 12 visual styles × 8 layouts × 3 color palettes

02~25 scene presets (knowledge-card / warning / poster / cozy-story, etc.) for one-click style + layout selection

03Image 1 anchor chain to ensure cross-card consistency (subsequent images reference Card 1)

04Three confirmation paths (quick confirmation / customization / detailed mode with three outline variants)

05Three reference image usage modes (direct / style / palette) without conflicting with the anchor chain

06Watermark support (appended to generation prompt when enabled in EXTEND.md)

07--yes non-interactive mode (skips all confirmations and initial setup, outputs directly using defaults, suitable for cron jobs)

08Style × layout compatibility warnings and automatic selection based on content signals

3External Dependencies

Kind	Dependency / Endpoint
api	运行时图像生成后端（Codex imagegen / Cursor GenerateImage / Hermes image_generate）——用户内容与提示词由此出网
cli	baoyu-image-gen 技能脚本（--provider codex-cli 首选路径）
cli	codex CLI（codex exec 后端，需 codex login）
cli	bun / npx -y bun（跑 baoyu-image-gen 或包装器）
network	homepage 元数据 URL（静态 frontmatter，任何步骤都不抓取）

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)

scripts/: Execution of local scripts or commands: This skill contains no executable scripts of its own; image generation is delegated by the agent to runtime backends (Codex imagegen Skill / Cursor GenerateImage, etc.).
网络端点: Declared external network access: The skill itself contains zero network code (no scripts or HTTP client imports in the directory).
本地工作目录: File I/O and artifact output: Writes image-cards/{topic-slug}/source-*.{ext} and analysis.md.

Audit Notes: According to the Batch 4 final ruling criteria (handing user content over to an external generation backend = orange), this skill provides two self-evident points: ① [Explicitly names runtime generation backends] The 'Image Generation Tools' section of SKILL.md names Codex `imagegen` ('If a skill named `imagegen` is listed, you are running inside Codex and MUST use it'), codex-imagegen ('**always** invoke through it rather than calling the wrapper directly.'), Cursor `GenerateImage`, and Hermes `image_generate`, defining resolution priority; consequently, user materials and card prompts (including direct --ref passes) are transmitted outbound to third-party models, consuming account quota, aligning with the platform's ai-image-generation orange classification. ② [Local disk writes] analysis.md, outline.md, prompts/NN-*.md, PNGs, and EXTEND.md are written to user directories with backup renaming rules. Why not yellow: Outbound activity involves generative service invocations rather than read-only queries to fixed endpoints. Why not red: No reverse engineering, no impersonation, no security downgrades (no sandbox bypass instructions), no credential reading, and zero executable scripts within the directory.

5Second-Pass Verification

[ok] 目录内无脚本、无网络代码 — 文件清单仅 SKILL.md + 15 份 references（presets/palettes/elements/workflows/config/codex-imagegen/style-presets/confirmation）；无 .ts/.sh/.py，无 http 客户端引用。
[ok] 后端解析阶梯与点名原文 — 四级阶梯（当前请求 > EXTEND.md > 运行时原生 > 唯一非原生 > 问用户）在 SKILL.md 顶部逐条列出，与 slide-deck 的对应段落逐字相同（同仓统一设计）。
[ok] 锚定链设计与后端限制自洽 — codex-imagegen.md 明写包装器不支持 --sessionId（'The wrapper does NOT accept a `--sessionId` flag.'），与 SKILL.md『链一致性必须来自 --ref』的说明一致。
[ok] 确认门禁与 --yes 出口 — Step 2 标为 REQUIRED 硬门禁，--yes 与『直接生成』类措辞可显式跳过，且跳过时要求先声明假定方案；非交互模式不写 EXTEND.md。
[ok] 首次设置为阻塞（与 slide-deck 相反） — first-time-setup.md 标 BLOCKING OPERATION，与 translate 同型，而与 slide-deck 的『不阻塞』形成对比；两处均在报告中如实记录。
[ok] 水印与写盘清单 — 水印提示词片段与 EXTEND.md watermark.* 字段对应；输出树 image-cards/{topic-slug}/ 与备份改名规则在 SKILL.md 明确写出。
[ok] 档位复核（批4 终裁） — 按 Main『点名生成后端 + 素材出网 → orange』落档；目录内无脚本无凭证，故不升红也不满足 green/blue 条件。
[ok] 元数据与 pin — 本地 HEAD = 1567581c26ec29f4216c6e6835415bf30343b0e3（2026-09-10 10:13:43 -0500）；GitHub API MIT / 25926 stars / pushed_at 2026-09-10T15:13:43Z；仓库根 LICENSE 为 MIT。

6Verdict & Scenarios

Visual system fully standardized and executable: combinations, presets, compatibility matrix, and auto-selection tables for 12 styles × 8 layouts × 3 palettes are all detailed in the main text.

Concrete mechanism for consistency rather than slogans: Image 1 anchor chain + layered reference images + sessionId (when supported by backend).

Reproducible and backend-switchable: prompt files strictly written first; image revisions require updating the file before regeneration.

Balances interaction and automation: Path A/B/C three confirmation tiers + full skip via --yes, fitting both manual fine-tuning and automated cron jobs.

Narrow attack surface: no scripts in the directory, no credential reading, no network access; all outbound requests occur via the user-selected backend.

Best for: Ideal for Xiaohongshu/WeChat article creators doing product recommendations and knowledge sharing: users who need visually consistent multi-card sets, want one-sentence generation while maintaining fine-tuning options, need to preserve a coherent series feel without design resources, and accept external image generation backend calls.

Not for: Not suitable for environments where asset transmission over the network is forbidden or data compliance is strictly sensitive; not suitable for materials requiring pixel-perfect text typography (programmatic patching is forbidden, requires regeneration); not suitable for environments lacking an available image backend; and not suitable for users expecting editable vector graphics (outputs bitmap cards).

Install Agent-Ready

① Verified Mirror 2026-09-15

Option A · Direct DownloadDownload baoyu-xhs-images.tar.gz

sha256: 4a9d014c6c3a9416…

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.