Design & Motion · JimLiu/baoyu-skills
baoyu-infographic
Generate professional infographics with 21 layout types and 22 visual styles. Analyzes content, recommends layout×style combinations, and generates publication-ready infographics. Use when user asks to create "infographic", "信息图", "visual summary", "可视化", or "高密度信息大图".
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
This directory contains zero scripts and zero executable code; all capabilities consist of procedural instructions in SKILL.md plus templates and knowledge bases in references/—making it a 'prompt engineering + reference library' skill.
skills/baoyu-infographic/SKILL.md
Two dimensions: **layout** (information structure) × **style** (visual aesthetics). Freely combine any layout with any style.
Directory contents are exclusively SKILL.md and references/ (21 layouts, 22 styles, base-prompt, analysis-framework, structured-content-template, two config items, and codex-imagegen instructions), without any .ts/.js/.sh files.
The recommendation algorithm operates on executable selection rules: content inference is bypassed to directly apply a mapped layout when keyword shortcuts match; otherwise, 3–5 layout×style combinations are recommended based on data structure, tone, and audience.
skills/baoyu-infographic/SKILL.md
Skip content-based layout inference for matched keywords.
Covers the two branches (3.1/3.2) in Step 3; SKILL.md also includes a fallback recommendation combination table (e.g., Timeline → linear-progression + craft-handmade).
Enforces a strict confirmation gate: Step 4 mandates user confirmation for the combination, aspect ratio, language, and backend, preventing the agent from generating images directly by default.
skills/baoyu-infographic/SKILL.md
**Hard gate**: this step is mandatory per the [Confirmation Policy](#confirmation-policy)
The Confirmation Policy section defines 'explicitly invoking the skill, supplying paths, matching keywords, or EXTEND.md defaults' strictly as recommendation inputs only, which do not authorize skipping confirmation; skipping requires explicit wording like --no-confirm or 'generate directly', and assumed values must be stated in the next user-visible message.
The image backend is resolved at runtime (native imagegen > codex CLI > Cursor native > other native > single non-native backend), explicitly naming baoyu-image-gen as a concrete route.
skills/baoyu-infographic/SKILL.md
route through `baoyu-image-gen --provider codex-cli` (preferred)
Mechanisms: pixel generation responsibility is delegated to host capability detection, but optional third-party skill routes are pinned at the command level; this serves as the skill's external dependency surface (user content and prompts outbound to third-party image models).
Prompt files are mandatory artifacts: the complete prompt must be saved to prompts/NN-{type}-[slug].md as a reproducible record before calling any backend.
skills/baoyu-infographic/SKILL.md
**Prompt file requirement (hard)**: write each image's full, final prompt to a standalone file under `prompts/` (naming: `NN-{type}-[slug].md`) BEFORE invoking any backend.
Step 5 synthesizes the prompt from layout definitions + style definitions + base-prompt template + Step 2 structured content + language; switching backends does not require rewriting the prompt.
Content fidelity is a strict rule: only transfer without summarizing, and credential-like information must be stripped prior to generation; the analysis phase employs an instructional design framework (learning objectives, cognitive load, data integrity).
skills/baoyu-infographic/SKILL.md
**strip any credentials, API keys, tokens, or secrets** before including in outputs
Core principle from the same section: `Preserve source data faithfully—no summarization or rephrasing`; references/analysis-framework.md further outlines 9 content-to-layout/style mappings alongside 'Data Integrity: Never alter, summarize, or paraphrase source facts'.
Templates use placeholders: base-prompt.md constructs the final prompt using {{LAYOUT}}/{{STYLE}}/{{ASPECT_RATIO}}/{{LANGUAGE}}/{{LAYOUT_GUIDELINES}}/{{STYLE_GUIDELINES}}/{{CONTENT}}/{{TEXT_LABELS}}.
skills/baoyu-infographic/references/base-prompt.md
- **Layout**: {{LAYOUT}}
The template encapsulates Image Specifications, Core Principles, and Text Requirements constraints, as well as `Generate the infographic based on the content below:` and the Text labels block.
Clear artifact and backup conventions: a fixed directory structure at infographic/{topic-slug}/, with timestamped backups for identical files and time suffixes appended during conflicts.
skills/baoyu-infographic/SKILL.md
infographic/{topic-slug}/
The Output Structure section specifies source-{slug}.{ext} / analysis.md / structured-content.md / prompts/infographic.md / infographic.png; Step 1.2, Step 5, and Step 6 each define backup renaming rules (e.g., `analysis-backup-YYYYMMDD-HHMMSS.md`).
Two strict 'no cheating' rules: masquerading raster graphics using SVG/HTML/canvas code rendering is prohibited, and painting over generated bitmaps with code to fix text is forbidden.
skills/baoyu-infographic/SKILL.md
**⛔ Never substitute SVG, HTML, canvas, or other code-based rendering for raster image generation.**
Rule two: `**⛔ Never repair rendered text by painting over a generated bitmap.**`—requires modifying the prompt to regenerate or switching layouts instead of applying image patches; post-processing is restricted exclusively to crop, resize, compress, and format conversion.
2Core Capabilities
01Orthogonal combination of 21 layouts × 22 styles (defaulting to bento-grid + craft-handmade)
02Keyword shortcuts (e.g., high-density information sheet → dense-modules + portrait)
03Content analysis → structured-content.md (verbatim preservation of title, learning objectives, sections, visual elements, text labels, and data points)
04Runtime image backend resolution and fallback chain (Codex imagegen / codex-imagegen / Cursor GenerateImage / other native / single non-native)
05Three reference image usage modes (direct pass to backend / style extraction / palette hex extraction) logged in prompt frontmatter
06Strict confirmation gate with explicit bypasses (--no-confirm / 'generate directly'), requiring assumed values to be declared when bypassed
07EXTEND.md preferences (preferred backend/layout/style/ratio/language/custom styles) and initial setup workflow
3External Dependencies
| Kind | Dependency / Endpoint |
|---|---|
| api | baoyu-image-gen（第三方 skill，--provider codex-cli 路径；会读该 skill 自己的 provider 凭据） |
| cli | Codex CLI（codex login 订阅态、codex exec） |
| api | Codex 运行时原生 imagegen skill（经由 Skill 工具调用） |
| api | Cursor 原生 GenerateImage 工具（无比例参数、无输出目录） |
| package | bundled codex-imagegen 包装脚本（免装 baoyu-image-gen 时的直调路径；其自身会 spawn codex，用 Codex 订阅而非 API key） |
| cli | bun（运行包装脚本；缺 bun 时 npx -y bun） |
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/: Execute local scripts or commands: Contains no native executable scripts; SKILL.md directs the agent to invoke image backends: runtime native imagegen / Cursor GenerateImage
- 网络端点: Declare external network access: Initiated by the resolved image backend (Codex imagegen / codex exec via Codex services; baoyu-image-gen via selected third-party services)
- 本地工作目录: File read/write and artifact output: infographic/{topic-slug}/: source-{slug}.{ext}, analysis.md, structured-content.md, prompts/, png
Audit Notes: Three-part justification: ① The evaluation target is this skill directory (SKILL.md + references/, containing no scripts, endpoints, or fetch calls); ② Directory-internal behavior consists entirely of local disk writes (source/analysis/structured-content/prompts/png + timestamped backups under infographic/{topic-slug}/); ③ The critical factor for the Amber rating is the external generation backend called for its own operation as named in SKILL.md: `route through baoyu-image-gen --provider codex-cli (preferred)` (alongside Codex native `imagegen`, Cursor `GenerateImage`, and bundled codex-imagegen). Under Batch 4 criteria, 'routing user material/prompts to external generation backends (involving account quota consumption and outbound data) → Amber'; credentials reside within the backend and this skill does not read keys, avoiding an escalation. No reverse engineering, impersonation, bypasses, or security downgrades are present (Note on details page: if the chosen backend is the codex-cli branch of baoyu-image-gen, it internally uses codex exec --sandbox danger-full-access, which is the backend's own risk profile; this skill does not alter files but notes it under verdict.risks).
5Second-Pass Verification
- [ok] 『21 布局 + 22 风格』数量属实 — references/layouts/ 实际 21 个 .md（linear-progression…dense-modules），references/styles/ 实际 22 个 .md（craft-handmade…retro-popup-pop）；SKILL.md 两张 Gallery 表逐项对应。
- [ok] 本目录无脚本/无端点/无凭证 — 目录内文件全为 .md；token 扫描 fetch/http/spawn/exec/process.env/API_KEY 零命中。
- [ok] 点名的外部生成后端（判橙依据） — SKILL.md Image Generation Tools 段逐条给出 Codex imagegen、codex-imagegen（含 `baoyu-image-gen --provider codex-cli` 首选路径）、Cursor GenerateImage、其他原生工具；references/codex-imagegen.md 进一步给出直调包装脚本的命令与 'no OPENAI_API_KEY is read or sent' 说明。
- [ok] 确认门禁为硬门禁而非建议 — SKILL.md 'Hard gate: this step is mandatory…' 与 Confirmation Policy 段把显式调用/关键词/EXTEND.md 默认值一律降级为 recommendation inputs；Step 6 亦注明必须先落 prompt 文件。
- [ok] 内容保真与脱敏要求 — SKILL.md Core Principles 原文 'Preserve source data faithfully—no summarization or rephrasing' 与 '**strip any credentials, API keys, tokens, or secrets** before including in outputs'；Step 2 Rules 重申。
- [ok] 反作弊条款（禁 SVG 替代 / 禁位图涂改） — 两段 ⛔ 文本均在 SKILL.md，且 Text correction policy 段限定后处理只能做裁切/缩放/压缩/转格式。
- [discrepancy] references 里 styles 定义的数量在 SKILL.md 自身文本中有笔误 — SKILL.md 末尾 References 列表写 'references/styles/<style>.md - 21 style definitions'，但 Style Gallery 与目录实际均为 22 个（frontmatter 描述也写 '22 visual styles'）。结论以目录/表格为准＝22；此为文档小笔误，不影响判级。
- [ok] codex-imagegen 直调路径可能导致下游红档行为 — 本 skill 的 references/codex-imagegen.md 指示在无 baoyu-image-gen 时直调 bundled 包装脚本，而该包装脚本（skills/baoyu-image-gen/scripts/codex-imagegen/spawn.ts）以 codex exec --sandbox danger-full-access 运行——红线属被调后端，改写为本 skill 的判级不成立，但已写进 verdict.risks 供用户知情。
6Verdict & Scenarios
Best for: Ideal for converting long-form articles, reports, or course materials into self-explanatory, dense infographics (especially 'high-density information sheets' via dense-modules + morandi-journal/pop-laboratory), as well as creators/educators needing consistent visual language who lock in defaults via EXTEND.md. Perfect for users who prioritize reviewing plans before image generation via strict confirmation gates.
Not for: Not suitable for workflows requiring pixel-perfect programmatic layout (which should use charting libraries or SVG code rendering, explicitly barred here as a substitute); environments lacking any usable image backend (halts at inquiry); un-sanitized transmission of confidential documents to external models; or batch production pipelines expecting single-pass perfection without multi-turn regeneration.
Install Agent-Ready
① Verified Mirror 2026-09-15
Option A · Direct DownloadDownload baoyu-infographic.tar.gz
sha256:
Option B · Copy Guide for Agent68d6840b2ed27261…Agents automatically download and verify mirrors from this JSON guide.
② Upstream GitHub
GitHub Repo ↗
Pinned to commit 1567581c26.