Design & Motion · helloianneo/ian-xiaohei-illustrations

ian-xiaohei-illustrations

Name: ian-xiaohei-illustrations
Author: helloianneo/ian-xiaohei-illustrations

Generates Ian-style body illustrations for Chinese text. Designed for requests to create "grotesque/quirky," "Xiaohei (Little Black)," "hand-drawn," "body illustrations," "article illustrations," "illustration suggestions," "shot lists," or "remove titles/edit images" for Chinese articles, posts, blogs, Notion docs, workflow guides, methodologies, processes, structures, states, metaphors, or viewpoints. By default, it adopts the Xiaohei character IP, pure white hand-drawn linework, minimal red/orange/blue annotations, and a clean, refreshing, yet imaginative visual style.

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

Positioned as a "Body Illustration Designer": first digests the article to locate cognitive anchors, drafts a shot list, and then generates 16:9 hand-drawn explanatory graphics one by one—neither generic illustration prompts nor PPT infographic templates.

ian-xiaohei-illustrations/SKILL.md

为中文文章设计和生成 16:9 横版正文配图。目标不是做商业插画、PPT 信息图或可爱卡通，而是把文章里的关键判断、流程、结构、状态或隐喻，变成一张清爽、怪诞、有创意、可读但不说明书的手绘解释图。

Visual IP is defined as a fixed character named "Xiaohei": solid black silhouette, white dot eyes, thin limbs, and a blank expression; must perform core actions within the frame rather than serving as mere decoration.

ian-xiaohei-illustrations/SKILL.md

黑色实心、白点眼、细腿、空表情，认真做一件荒诞但成立的事。

Paired with rigid style-dna requirements: pure white background, black hand-drawn line art, subject occupying 40%-60%, at least 35% negative space, and a maximum of 5-8 Chinese annotations.

Image generation is explicitly delegated to the host's built-in tool (image_gen) and must be executed one by one without collages; it does not wait for secondary confirmation when the user explicitly requests generation.

ian-xiaohei-illustrations/SKILL.md

不要停下来等确认；用内置 `image_gen` 每张单独生成。不要把多张图拼在一张里。

Critical grading factor: SKILL.md names the host runtime tool used for image generation; execution sends article content and prompts to the image generation backend (consuming account credits), placing it in orange rather than blue/green.

Prompts follow a fixed framework (prompt template file), populating style DNA, IP character, structural pattern, core concept, composition, visual elements, Chinese labels, color palette constraints, and negative constraints step by step.

ian-xiaohei-illustrations/references/prompt-template.md

每张图单独生成。根据正文内容替换变量，不要把多张图拼在一起。

The template includes replaceable variables (topic / structure type / core meaning / specific scene / elements / labels) and an English Visual DNA constraint block.

Anti-templatization discipline: explicitly forbids replicating layouts from past cases, requiring every article to reinvent metaphors from scratch; sample images serve only for low-frequency visual calibration.

ian-xiaohei-illustrations/SKILL.md

不要复刻过往案例。案例只提供风格密度和小黑参与方式，不能直接复用“传送带断点 / 小黑拉线 / 素材鱼 / 盖章工具箱 / 常见坑路径”等已有构图，除非用户明确要求复刻某张图。每次都要从当前文章重新发明一个奇怪但成立的隐喻。

Executes a pre-delivery QA checklist; failure signals include Xiaohei serving only as decoration, overcrowded frames, looking like PPTs/flowcharts, category titles in the top-left corner, Chinese typos, overly cute art style, and dirty backgrounds.

ian-xiaohei-illustrations/references/qa-checklist.md

## 失败信号出现以下情况，重生成或局部编辑： - 左上角有“常见坑 / Workflow / 系统架构图 / 路线图”等标题。 - 小黑像吉祥物、表情包或可爱卡通。 - 画面像 PPT、课程课件、正式流程图。

The illustration strategy first outputs a shot list (specifying placement, theme, core meaning, structure type, Xiaohei's action, suggested elements, and label words for each image) with strict quantity caps: 4-8 images by default, 1-3 for short texts, and no more than 9 for long texts.

ian-xiaohei-illustrations/SKILL.md

先给 shot list。每张图写清楚： - 放在哪个段落后 - 图的主题 - 核心意思 - 结构类型 - 小黑在图里做什么 - 建议元素 - 建议中文标注词

The text-image relationship is treated as a hard requirement: select only cognitive anchors (core judgments, two breakpoints, input-output loops, branching paths, before-after contrasts, multi-purposing, transition paths, common pitfalls, character state transitions) instead of distributing illustrations evenly.

ian-xiaohei-illustrations/SKILL.md

不要平均配图。优先选择

Delivery must persist to files: copy final images to assets/<article-slug>-illustrations/ and name them sequentially with 01-/02- prefixes, preserving original generated files without overwriting existing assets.

ian-xiaohei-illustrations/SKILL.md

assets/<article-slug>-illustrations/

2Core Capabilities

01Article illustration strategy and shot list planning (placement, subject, structural pattern, Xiaohei's actions)

02Sequential generation of 16:9 body illustrations in Xiaohei IP style (via host built-in image_gen)

03Structured prompt templates (style DNA / IP / structural pattern / composition / annotations / negative constraints)

04Localized prompt editing: removing top-left titles, boosting quirky/grotesque feel (while keeping the rest intact)

05Post-generation QA and regeneration criteria (decorative use, overcrowding, PPT feel, typos, over-cuteness, dirty background)

06Style DNA constraints (pure white background, hand-drawn lines, negative space ratio, color semantics: orange=primary path, red=focus/issues, blue=supplementary/feedback)

07File delivery standards (assets/<slug>-illustrations/, sequential prefix naming, never overwriting existing assets)

08Codex host manifest (display_name, short_description, default prompt, implicit invocation allowed)

3External Dependencies

Kind	Dependency / Endpoint
api	宿主内置图像生成工具 image_gen（出图后端由宿主决定）

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)

网络端点: Declared external network access: No networking code or endpoints within the skill directory; image generation calls are handled via the host's built-in image_gen tool (prompts and article content reach the image generation backend, provider determined by the host; this sk...
本地工作目录: File read/write and artifact output: Copies final images into `assets/<article-slug>-illustrations/` named with 01-/02- prefixes; requires preserving original generated files without...
SKILL.md: Declared invocation of host CLI interpreter or external execution environment

Audit Notes: The evaluation target is this skill directory (SKILL.md + 5 references + agents/openai.yaml + assets/examples sample images). The grading rationale is "sending user materials/prompts to an external generation backend": SKILL.md explicitly requires "generating each image individually with the built-in `image_gen`"; upon execution, article content and prompts are sent to image generation models (consuming account credits, data leaves local environment)—sharing the same generation backend category as site skills like ai-image-generation / baoyu-article-illustrator. The skill itself contains zero scripts, zero endpoints, zero credential reads, and zero security downgrades, so it is not red; local side effects are strictly limited to copying images to assets/<slug>-illustrations/.

5Second-Pass Verification

[ok] skill 目录无脚本、无端点 — 目录内只有 6 个 markdown/yaml 与示例 PNG；URL/curl/requests 扫描零命中；出图完全依赖宿主 image_gen。
[ok] image_gen 点名（判橙依据） — 原文「如果用户明确要求「生成 / 输出 / 做图 / 帮我生成」，不要停下来等确认；用内置 `image_gen` 每张单独生成。不要把多张图拼在一张里。」——工具被点名，执行后素材出网。
[ok] 无凭证读取 — API_KEY/token/secret/keychain 零命中；agent 清单里无 env/凭证字段。
[ok] 写盘范围 — 只要求把最终图复制到 `assets/<article-slug>-illustrations/`（放在用户 workspace 内），并「保留原始生成文件，不要覆盖已有资产，除非用户明确要求替换」。
[ok] 反复刻纪律 — 「不要复刻过往案例。案例只提供风格密度和小黑参与方式…除非用户明确要求复刻某张图。每次都要从当前文章重新发明一个奇怪但成立的隐喻。」逐字核对。
[ok] IP 强约束 — 「小黑必须参与画面的核心动作，不能只是站在旁边当装饰。」+ QA 必过项「小黑承担核心动作，不只是装饰。」两处一致。
[ok] 示例图的定位 — SKILL.md「assets/examples/：只作低频视觉校准，不进入默认生成路径。不要照抄这些案例的构图、物件或标注。」；NOTICE.md 说明示例图由 Ian 生成并署名。
[ok] 数量与粒度限制 — 「默认 4-8 张。文章很短时 1-3 张；长文也不要轻易超过 9 张。够用就好，避免把正文做成画册。」+「每张图只讲一个核心结构」。

6Verdict & Scenarios

Personified IP + rigid visual constraints (pure white background, hand-drawn lines, negative space ratio, color semantics) yield high recognizability without degrading into generic AI art.

Clear anti-templatization discipline: bans copying past cases, bans top-left type headers, and bans PPT or overly cute styles, turning 'not looking like generic AI art' into verifiable checklist items.

Complete and reproducible workflow: Strategy (shot list) → Sequential prompt formulation → Local editing → QA → Disk persistence and naming, with templates provided for each step.

Quantity control and 'one structure per image' principle prevent turning the article into a crowded picture book.

Clean attribution and licensing handling (Xiaohei IP + example image credits + author links).

Best for: Ideal for long-form Chinese content creators needing batch article illustrations: users who need consistent, recognizable, white-background hand-drawn explanatory graphics featuring a character IP, and are comfortable with generative image creation (materials leave the local environment, requiring multi-turn iterations). Also fits as a reusable plug-and-play toolkit ('illustration strategy + prompt templates + QA checklist') across other image models.

Not for: Not suitable for scenarios sensitive to external data transmission (drafts are sent to image generation backends); not suitable for tasks requiring precise infographics, charts, or data visualizations (strictly rejects PPT and flowchart styles); not suitable for workflows demanding one-shot zero-proofreading generation (Chinese labels and IP consistency require iteration); not intended for commercial advertising illustrations or brand visual identity deliveries—its core positioning is strictly 'body text illustrations'.

Install Agent-Ready

① Verified Mirror 2026-09-15

Option A · Direct DownloadDownload ian-xiaohei-illustrations.tar.gz

sha256: 88e4f6c95f21cc30…

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 91b560849e.