Design & Motion · wnby/photo-relic-editorial

photo-relic-editorial

Name: photo-relic-editorial
Author: wnby/photo-relic-editorial

A Codex skill for transforming real photos into Photo Relic artworks: truthful photographs paired with quiet paper-memory prints.

L1 Basic · Source registration onlyV1 Commit-PinnedDual-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

Zero-code dual-layer pure prompt implementation combining orchestration and knowledge base: SKILL.md serves as the workflow and aesthetic discipline layer (inspect photo → extract cues → select recipe → generate → run QA), while references/afterimage-editorial-prompt.md acts as a prompt compiler that SKILL.md mandates reading first (structural constraints, fidelity bans, subject treatments, signature system, and ready-to-use Prompt Templates).

SKILL.md

Read `references/afterimage-editorial-prompt.md` before composing the final image prompt.

This is the core engine of the skill: the true craft—how to write photo fidelity constraints, how to translate architecture, skylines, and figures into printmaking marks, and template syntax—is entirely codified within references documents, which SKILL.md enforces invoking before generation.

Pixel generation is fully delegated to the host's image tool, with the skill itself containing no scripts, API endpoints, or vendor lock-in: SKILL.md only mandates using the image generation/editing tool with the user's photo as the reference image, leaving vendor choice to the host (Codex's image tool).

SKILL.md

This is a creative image-generation skill. When producing the final image, use the image generation/editing tool with the user's photo as the reference image.

C2PA metadata in the bundled sample PNG reveals software agent strings `nameigpt-imagegversionc2.0` and `c2pa.watermarked.unbound`, indicating the author originally generated outputs using OpenAI gpt-image 2.0. However, this fact exists only in sample artifact metadata; the skill instructions do not name any vendor, allowing seamless portability across hosts and image tools.

The core creative constraint is the dichotomy between photo-area truth value and lower relic-area abstraction: any redrawing, beautifying, outpainting, or stylization is strictly forbidden on the real photograph area, confining abstraction to the lower relic area to ensure the final output is a true photo paired with a memory print rather than an all-over filter.

references/afterimage-editorial-prompt.md

Use abstraction only in the relic region unless the user explicitly asks for an integrated edit.

SKILL.md reinforces this with stringent wording: 'Preserve the photograph's content and truth. Do not redraw, beautify, repaint, expand, hallucinate, or stylize the original photographic area.' The photo is treated as immutable proof, matching the reference doc's verbatim principle: 'The photograph is evidence.'

Aesthetic decisions are parameterized into a 5-axis recipe system (layout rhythm / relic grammar / mark weight / title mode / motion seed) with discrete choices, requiring selection before prompt assembly to guarantee cross-photo series consistency while introducing controlled variations.

SKILL.md

Pick one option from each axis before writing the image prompt. Vary the recipe when recent outputs look too similar, but keep the series identity stable.

The Prompt Compiler section in references mandates embedding the recipe directly into a fixed slot within the final prompt: 'Use this recipe: [layout rhythm] / [relic grammar] / [mark weight] / [title mode] / [motion seed]'—making the recipe part of the generation instruction rather than internal jargon.

A translation table mapping details into marks serves as the direct source of rendering quality: roofs → overlapping arcs, windows → sparse cuts/dots, people → short vertical incisions, water → one or two horizontal residue lines, preventing the panel from turning into an illustration or decoration.

SKILL.md

Translate details into marks, not decorations: roof layers become stacked arcs; windows become sparse cuts or tiny dots; people become short vertical ticks; water becomes one or two horizontal residues; dusk becomes one small warm signal.

The Subject Treatments section in references defines consistent translation rules by subject (architectural landmarks, urban skylines, landscapes/roads/plazas, and figures), such as rendering humans strictly as 'small irregular vertical marks' while forbidding faces, limbs, or clothing details.

Dual Quality Gates (ten checks in SKILL.md + nine in references) perform checklist-based self-verification before generation; any primary failure permits only a single, shorter, and stricter regeneration, preventing mindless retries or lowering passing standards.

SKILL.md

If one major item fails, regenerate once with a shorter, stricter prompt focused on that failure.

The references gate provides concrete examples of corrective phrasing: 'make the temple relic more like a printmaking seal and less like watercolor' / 'increase subject recognition while preserving generous blank paper,' translating failure modes directly into actionable correction prompts.

Codex ecosystem entry point: agents/openai.yaml provides display_name, short_description, and default_prompt; the README guides copying the entire directory into ~/.codex/skills and triggering it via $photo-relic-editorial, while the Chinese README demonstrates a serialized four-character Chinese title pattern.

agents/openai.yaml

default_prompt: "Use $photo-relic-editorial to turn this photo into a Photo Relic artwork with a truthful photo region and a recognizable abstract memory print."

The README installation section states: 'Copy the entire `photo-relic-editorial` folder into your Codex skills directory.' Example invocation: 'Use $photo-relic-editorial to turn this photo into a Photo Relic artwork.' The skill relies on the user actively attaching a photo, permitting requests only when missing: 'Ask for the missing photo only if no usable source image is available.'

2Core Capabilities

01Generates vertical dual-zone editorial artwork combining an authentic photograph and a paper memory print (photo relic editorial layout)

02Pre-generation source image reconnaissance: extracts 3–5 categories of cues from real photos (subject relationships, mood concept, primary/accent colors, light direction, structural vocabulary) as the sole material source for the relic

03Photo-zone truth preservation: strictly forbids redrawing, beautifying, repainting, outpainting, hallucinating, or stylizing the photo zone, ensuring the photograph remains recognizable in its original state

04Five-axis recipe system (layout rhythm, relic grammar, mark weight, title mode, motion seed) offering discrete menu-style combinations to maintain series consistency with controlled variation

05Subject-specific translation manual: concrete rules and negative constraints for reducing architecture, skylines, natural landscapes, and crowds into print marks

06Restrained typographical discipline: minimal, poetic, caption-like titling that avoids commercial advertising or poster aesthetics

07Social/short-video (Douyin) readiness: 9:16 cover layout and motion seeds allowing static compositions to transition smoothly into an animation sequence: photo → silhouette drop → mark assembly → final title emergence

08Pre-generation quality gates: a ten-item checklist (photo truth, thumbnail legibility, intact subject relationships, series signature, palette provenance, no UI watermarks, etc.) allowing only one stricter regeneration upon failure

3External Dependencies

Kind	Dependency / Endpoint
api	宿主图像生成/编辑工具（参考图输入；SKILL.md 未点名厂商，供应商由 Codex 宿主决定）

4Recommended Governance & Evidence

Suggested Controls for Safe Operation

For Agent Runtime

Can be auto-loaded

For Administrators

Evidence & Triggers (File Paths & Excerpts)

SKILL.md: Pure prompt instructions and rule text, without script execution or external impacts

Audit Notes: Purely instruction- and knowledge-based skill: The repository contains no executable scripts (no .sh, .py, or binary code; the entire repo consists strictly of md, yaml, png, and LICENSE files). Token scans across text files yield zero hits for URL, curl, wget, API keys, tokens, secrets, .env, bash, or exec commands—no script execution, no skill-level network egress, no credential reading, and no file-write instructions. The sole execution path is the host's built-in image generation/editing tool (SKILL.md explicitly references 'the image generation/editing tool' without specifying vendor endpoints); user photos are forwarded to image models through host tools upon user request, which is an intrinsic functional behavior governed by the host. The .gitignore reservation of generated/ and scratch/ directories implies output persistence is managed entirely by the host. Key operational notes: ① Source materials require proper user rights (declared in the README MIT section); ② Sample images are AI-generated aesthetic anchors rather than real photographs; ③ Actual quota usage and privacy policies depend on the host and model vendor, entirely outside the skill's reach.

5Second-Pass Verification

[ok] 外部依赖唯一条目：宿主 image tool + 无厂商端点 — SKILL.md 全文本扫描确认只有 'use the image generation/editing tool' 一类泛称，无 openai/gpt/anthropic/replicate 等点名；示例 PNG 的 C2PA 元数据（gpt-image 2.0）属产物证据而非指令，已在条目 note 中如实标注为『实际出图工具』推断。
[ok] security.grade=green 的反例搜查（隐藏脚本/网络/凭证） — 隐藏文件仅 .gitignore/.DS_Store 类忽略项；文本文件 URL/token/exec 模式零命中；无 package.json/requirements/*.sh；唯一网络痕迹出现在 examples/*.png 的 C2PA 证书链（cv.iptc.org、ocsp-c2pa.ssl.com、trufo.ai），属图片内嵌内容凭证，非可执行代码。
[ok] 『照片区不可篡改』声明 vs 实际能力 — 该声明是写给宿主图像模型的文本禁令（SKILL.md Creative Rules + references Photo Preservation 双重强调），skill 无代码可强制执行；第一遍已将其表述为『prompt 层约束 + Quality Gate 兜底』，未夸大为硬保证。
[discrepancy] README 'Paper Beijing' 示例图身份 — 第一遍初读时按 README 语境默认示例是真实照片经 skill 处理的成图——成立；但第二遍在 PNG 内发现 C2PA 软件代理串 'nameigpt-imagegversionc2.0' 与 'c2pa.watermarked.unbound'，确认示例成图由 OpenAI gpt-image 2.0 生成并含水印/内容凭证动作。此发现不推翻结论，但 internal_assets 与 verdict 对示例图的描述已按此修正（AI 生成示例，作为审美锚点）。
[ok] references 文档『Prompt Compiler / Quality Gate』引用准确性 — 逐字重读 156 行全文，第一遍所用 quote（recipe 槽位模板、'If a major gate fails, regenerate once with a shorter corrective prompt…'、'Use abstraction only in the relic region…'、Signature System 色板/比例）均与原文一致。
[ok] 元数据：commit/license/stars/last_push — GitHub API：spdx mit、stargazers_count 378、pushed_at 2026-08-05T02:37:31Z（= pin commit 时间，仓库 3 个 commit 无更新）；本地镜像 HEAD == pin sha 2232da16afddc7940e2e2f280bfb85aa62da1bae。

6Verdict & Scenarios

Transforms the aesthetic system of turning photos into art into a reproducible engineering pipeline (cue extraction → recipe selection → translation table → quality gate), enabling beginners to reliably produce unified series without relying on random model luck.

Photo truth ethics serve as a distinctive differentiator: strictly forbidding redrawing, beautifying, or hallucinating the photo zone and confining abstraction to the relic zone aligns with compliance narratives around creating derivative art from licensed photos (README License section).

Prompt output discipline (separating analysis from generation instructions and outputting production language only) paired with a self-limiting single stricter retry ensures clean prompts and bounded failure recovery.

Zero-code and zero dependencies, introducing no attack surface beyond standard installation, fully portable across hosts (runs on any agent with image editing tools), and instantly ready for Codex (openai.yaml + $photo-relic-editorial).

Best for: Ideal for users looking to transform travel, architectural, or urban photography into reusable vertical art series (especially well-suited for historic architecture, landmarks, skylines, and scenes with strong light, shadow, and order); tailored for Codex ecosystem creators aiming to produce dual-zone 'real photo + printmaking memory' compositions with headroom for Douyin 9:16 short-video expansion; also serves as an exemplary reference for skill authors designing dual-layer 'prompt orchestration + references knowledge base' architectures and photo ethical constraint writing.

Not for: Not suitable for image retouching workflows requiring fine local compositing, multi-image blending, or redrawing the original photo (the skill strictly forbids modifying the photo zone); not suitable for high-speed bulk generation where stylistic consistency is ignored (its value lies in restrained series signatures); not suitable for host environments lacking image generation/editing tools (no local fallback rendering path); not suitable for workflows with strict vendor privacy constraints where photos cannot leave the local network.

Install Agent-Ready

① Verified Mirror 2026-09-09

Option A · Direct DownloadDownload photo-relic-editorial.tar.gz

sha256: 38c0459aca3ab84d…

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 2232da16af.