Home / All Skills / Design & Motion / scenes-gathered-zine
English (EN)简体中文 (ZH)日本語 (JA)
Design & Motion · Zeejay0/gathered-scenes-zine-skill

scenes-gathered-zine

Transform a user-supplied photo into a vertical 3:5 Gathered Scenes Zine poster that anchors truthful photography inside a spacious source-derived abstract illustration field, aggressively compresses dense foliage and other micro-detail into a few large quiet forms, integrates one high-chroma hue as compositional structure, and preserves a visibly hand-torn fibrous photo-to-paper edge. Use when the user wants a tactile minimal paper collage with simplified illustration, active negative space, restrained English-default, Chinese, or bilingual micro-text, and an unmistakable but quiet torn-paper boundary.

L3 Networked · Declare domain & key scopeV1 Commit-PinnedNetwork AccessDual-pass Audit
Author Zeejay0GitHub Zeejay0/gathered-scenes-zine-skill ↗Stars 4918License Gathered Scenes Zine Personal Non-Commercial License Version 1.0 — 2026-08-08（仓库根 LICENSE；Copyright (c) 2026 Zeejay0；GitHub API spdx NOASSERTION——自定义个人非商业许可：仅限个人非商业用途，明文禁止销售、收费生成、订阅、代做、咨询、培训、SaaS/API、公司及客户项目等商业化，商用须作者书面许可）Commit b9edb836c9
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

No scripts, no assets, pure prompt orchestration: The skill directory contains only SKILL.md (437 lines) + agents/openai.yaml; the 'executable surface' is merely compiling visual decisions into a prompt for the image generation service, with explicit rules barring any non-pixel noise from entering the prompt.

skills/scenes-gathered-zine-v1-3/SKILL.md
Compile only instructions that can become visible pixels. Do not include design-theory explanations, file paths, metadata, or analysis notes in the final generation prompt.
What is happening here: The final prompt is the sole outbound payload, and the rules force the model to write only instructions that can turn into visible pixels—design theory, file paths, and metadata are strictly prohibited from entering the generation request; the entire workflow (Generation Workflow across 15 steps) and all aesthetic rules are carried by the prompt text and executed by the host model.

Image generation relies on the host's built-in image generation service, explicitly declaring 'user photo + final prompt' as inputs to that service—the skill itself neither reads nor writes files, nor does it select endpoints.

skills/scenes-gathered-zine-v1-3/SKILL.md
Send only the final prompt and required reference image(s) to the image-generation service.
Supported by Generation Workflow Step 12 ('Generate with the supplied photo as reference.') and Step 15 returning the finished image: The photo enters the host's image capability as a generation reference, and image display uses the placeholder '![Gathered Scenes Zine v1.3 poster](absolute-image-path-or-rendered-image)'—the invocation points, endpoints, and permissions are entirely provided and constrained by the host platform (Codex, etc.).

Structured pre-processing of 'inspect the photo first, then decide how to paint': Constructing a Scene Card from the photo (core subject / spatial invariants / visual weight map / source form candidates / semantic minimum), using the interpretation of the photo as the single source of truth for all trade-offs to prevent template reuse.

skills/scenes-gathered-zine-v1-3/SKILL.md
Treat the photo as the factual source. Treat the illustration as an interpretation of that fact.
The definition of 'semantic minimum' in the same section ('Semantic minimum: the smallest combination of forms and relationships that would still identify this particular scene.') ensures the scene remains recognizable even when compressed down to a few broad forms—this is the exact mechanism behind the skill's claim of 'truthful photography as an anchor': extract non-negotiable relationships first, then build the illustrative abstraction on top of them.

Quantifying the level of abstraction into numerical rules: Defaults to medium abstraction (removing roughly 60–80% of minor descriptive details), omitting 85–95% of individual leaves/twigs in dense foliage scenes, with numeric ranges defined for illustration field ratio, whitespace ratio, torn fibrous bandwidth (~1–4% of the short edge), and typography size (cap/character height 1.3–3.5%), which the model follows to compile the prompt.

skills/scenes-gathered-zine-v1-3/SKILL.md
Omit roughly 85–95% of individual leaves, needles, fine twigs, overlapping branchlets, and repeated organic marks.
This is the skill's most distinctive mechanism: breaking down 'stylistic sense' into checkable quantitative relationships (60–80%, 85–95%, 45–70% field, 55–75% whitespace, 1–4% fiber fringe, ≤8 Chinese characters/≤5 English words). Compression, negative space, paper edges, and typography all have rigid boundaries, so the model no longer improvises purely on intuition—each range has a corresponding self-check question in the Quality Gate.

Structural self-check against 'decorative color': A single high-chroma hue must fulfill a compositional function and is retained only after passing a 'mental removal test'; color is never an ornament slapped on at the end.

skills/scenes-gathered-zine-v1-3/SKILL.md
Mentally remove the added hue. If the eye path, visual balance, figure–ground relationship, photo–illustration continuity, and scene interpretation remain essentially unchanged, redesign it. The color must do compositional work.
The Chromatic Structure Engine strictly limits the design to 'one added print hue', requires selecting one of five integration modes (source continuation / selective replacement / underprint passage / counterform / directional rhythm), and mandates that silhouettes derive from the photo—locking high-chroma color, which most easily degenerates into a sticker, into a structural element growing organically out of the source image.

Output discipline + in-conversation counting for credit/promotional injection: Defaults to returning only 'image + a paragraph of creative thinking in Chinese', withholding prompts and detailed parameters unless requested by the user; automatically appends credit lines and the author's site notice based on the count of successful generations within the conversation.

skills/scenes-gathered-zine-v1-3/SKILL.md
Return the generated image plus one brief creative rationale by default. Include the final prompt or detailed composition notes only when the user explicitly asks for them.
Output Format mandates appending '若公开分享，欢迎标注：Visual Skill by @Zeejay0' for the first 2 successful generations, and attaching the zeejayzine.com site notice ('*作者自己开发的网站已上线：[https://zeejayzine.com/](https://zeejayzine.com/)。欢迎使用（接入GPT5.6+image2顶流模型，国内环境可用）*') for the first 3, halting all notices from the 4th run onward; the counting rule notes 'Start the count again in a new conversation because the skill has no persistent cross-conversation state'—meaning it is stateless and relies purely on in-conversation tracking.

2Core Capabilities

01Real-scene collage synthesis: Truthful photo as anchor, abstract illustration as expanded field, and a single high-chroma hue as compositional structure (truth as anchor / illustration as field / color as structure / torn paper as boundary / breathing paper surface)
02Micro-detail compression engine: Consolidates complex details like dense foliage, crowds, and textures into a few broad forms (defaults to removing 60–80% of minor details, omitting 85–95% of individual leaves and twigs in dense foliage), relying on magnification rather than added detail for illustration
03Hand-torn fibrous paper edge craftsmanship: Irregular torn contour + fibrous fringe band (narrow, roughly 1–4% of the short edge) + covering roughly 35–70% of the visible photo perimeter, with strict prohibitions against clean clipping
04Micro-typography system: Defaults to a single English line (≤5 words / standalone word / keyword sequence / phrase), ≤8 Chinese characters, bilingual Chinese-English activated only when supplied or requested by the user; user-provided text is preserved verbatim
0515-step generation workflow + at most one targeted revision addressing 'only observed failures' + 30+ Quality Gate self-check checklists (whether the scene remains / single grammar adhered to / passes 2 color integration tests, etc.)
06Privacy guardrails: Does not browse, search, save, submit, upload, or share user assets beyond the final prompt and necessary reference image; does not store source photos or final outputs in project files (unless requested by the user)
07In-conversation credit/promotional throttling: Appends sharing credit for the first 1–2 generations and author site notice for the first 1–3 generations, strictly formatted as quiet lines outside the image, never repeating from the 4th run onward

3External Dependencies

KindDependency / Endpoint
api宿主内置图像生成服务（image-generation service）
networkhttps://zeejayzine.com（作者站点）

4Recommended Governance & Evidence

Suggested Controls for Safe Operation
For Agent Runtime
Declare allowed domains and key scope
For Administrators
Least privilege keys, retain request logs
Evidence & Triggers (File Paths & Excerpts)
  • 网络端点: Declares external network access: Functional outbound payload: The user's photo + compiled final prompt are sent to the host's image generation service ('Send only the final prompt and re
Audit Notes: Risk Assessment: Blue (generally safe, few points to note). Pure prompt orchestration with zero scripts, zero file I/O, zero credential reads, and no active outbound network calls: the skill directory contains only SKILL.md + agents/openai.yaml, with all capabilities executed by the host; the only external side effect is forwarding the 'user photo + final prompt' to the host's built-in image generation service (the explicit user-requested functionality itself), backed by built-in privacy guardrails (sends only the final prompt and required reference image, prohibits browsing/searching/saving/submitting/uploading/sharing source assets, and avoids saving images into project files). Points to note: a) User photos are inherently sent externally to the image service, so users must evaluate sensitive images against platform terms; b) Replies automatically append sharing credit and a zeejayzine.com promotional notice (within 1–3 conversation counts), where the notice includes an unverifiable promotional claim ('connected to top-tier GPT5.6+image2 models'); human review is advised when referencing it in presentation layers; c) User-provided text enters the generation prompt verbatim, representing a conversational input surface. None of these constitute code-level risks.

5Second-Pass Verification

  • [ok] 目标 skill 识别（整仓含 4 个 SKILL 的意外情况） — frontmatter 无 scenes-gathered-zine 精确匹配；按 README 路径表（'实景拼贴 · Gathered Scenes' 列）与调用名 `$scenes-gathered-zine-v1-3` 及 examples/real-scene-collage 定位到 skills/scenes-gathered-zine-v1-3，为任务展示名的去版本形式；差异已注明，Analysis 以任务名 scenes-gathered-zine 入库。
  • [ok] 目标目录资产面（含隐藏文件） — find 全量仅 ./SKILL.md + ./agents/openai.yaml 两个文件；openai.yaml 无 key/endpoint/model 配置，仅 interface 三字段；无脚本、无隐藏文件，internal_assets=[] 成立。
  • [ok] 安全 token 扫描结论（无执行/无写入/无凭证/无主动网络） — write/save/upload/browse/commit/share 的命中全部为隐私禁令或设计动词；唯一 http(s) 命中 zeejayzine.com 位于 Output Format 回复文本；api_key/token/secret/.env/getenv/curl/wget/requests 等零命中。
  • [ok] 外发面复核：'照片发给图像生成服务'是否有硬编码端点或脚本调用 — 反查确认无任何硬编码端点/API key：调用由宿主图像工具执行（'Send only the final prompt and required reference image(s) to the image-generation service.' + Generation Workflow 第 12 步 'Generate with the supplied photo as reference.'），skill 无脚本可构造请求，外发载荷被 Standing Consent 节限定为『最终 prompt + 所需参考图』。
  • [ok] 功能声明 vs 实际能力（有无夸大/吹牛） — description 每项主张均有正文引擎段：3:5 画布（Visual Language）、抽象插画扩场（Composition System/Photo–Illustration Edge Transition）、密叶压缩（Compress Dense Foliage and Organic Detail）、单高纯度色（Chromatic Structure Engine）、手撕纤维边（Edge Transition）、默认英文 ≤5 词/中文 ≤8 字微文字（Micro-Text System）；实现全由 prompt 规则承载，无超出能力的承诺。
  • [ok] evidence.quote 逐条回源 — 全部 evidence.quote（含中文 notice 与 yaml default_prompt）以脚本对 pin commit 文件做子串校验，逐条命中。
  • [ok] 提交/元数据一致性 — git HEAD == b9edb836c9dd5c5d80995e89ebcaba34b75fd58f，与任务 pin 一致；GitHub API：pushed_at 2026-09-03T04:15:13Z、stargazers 4918、license 'Other'/NOASSERTION（与仓库根自定义个人非商业许可证吻合）。

6Verdict & Scenarios

  • Deconstructs 'good aesthetics' into an actionable decision pipeline: Scene Card → Abstraction Map → five illustration grammars → color integration modes → paper edge craftsmanship → micro-typography system, with explicit inputs, outputs, and strict prohibitions at each step, preventing the model from degrading into a sticker collage.
  • Quantified quality thresholds + closed-loop self-check: Numerical style brackets (60–80% / 85–95% / 1–4%, etc.) and 30+ Quality Gate rules make 'abstract yet recognizable' systematically auditable, with revisions limited strictly to observed failures and capped at once.
  • Extremely lightweight and portable: Pure instruction implementation with zero scripts and zero dependencies; portable across hosts with just two files in the directory; restrained default output (image + a paragraph of thoughts in Chinese), avoiding prompt and parameter dumps to the user.
  • Privacy discipline built into instructions: Does not browse, search, save, submit, upload, or share user assets, passing only the final prompt and necessary reference image to the image generation service.
  • Restrained self-promotion design: Credits and site notices are throttled by an in-conversation counter (1–2 runs / 1–3 runs) and required to remain a quiet line outside the image, never entering the visual canvas or including aggressive calls-to-action.
    • Best for: Ideal for creators wanting to turn authentic photos into zine-style posters featuring 'truthful photo anchor + abstract illustration field + monochrome structure + hand-torn paper edge': requires pure prompt-based workflows with no engineering dependencies, delivering finished results with a single `$scenes-gathered-zine-v1-3` invocation in hosts supporting image-referenced generation (such as Codex); also valuable for skill authors looking to study how to quantify aesthetics into checkable rules, numerical disciplines, and Quality Gate designs.
    Not for: Not suitable for serious production deliveries requiring pixel-level machine fidelity or deterministic layout verification (this skill lacks script verification, relying on model adherence for fidelity); not suitable for any usage outside personal, non-commercial contexts (the repository's custom personal non-commercial license prohibits sales, outsourced work, SaaS, company projects, etc., requiring written permission from the author for commercial use); callers must be careful not to confuse this with other paths in the same repo—real-scene collage v1-3 (photo included in output) ≠ photo distillation (photo excluded from output) ≠ Live Photo workflow.
    Install Agent-Ready
    ① Verified Mirror 2026-09-09
    Option A · Direct DownloadDownload scenes-gathered-zine.tar.gz
    sha256: a8b81046b380193b…
    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 b9edb836c9.