技能 relight

📦

relight

Name: relight
Author: prime-skills

內容修訂 r1 安全 ⚙️ 外部命令🌐 網路存取📁 檔案系統存取

Relight Still Images with RunComfy

Existing photos may have lighting that does not match the desired mood or product presentation. This skill guides RunComfy relighting with model selection and precise lighting prompts.

支援: Claude Codex Code(CC)

🥉 77 銅級

Maintain this Skill? Claim it 🥉 77 銅級

使用我的智慧代理安裝

將此請求複製給你的智慧代理，其中包含權威 Skill 頁面和清單。

智能體請求

Review the Skillstore skill "relight" from https://skillstore.io/skills/prime-skills-relight.md and its manifest at https://skillstore.io/api/skills/prime-skills-relight/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.

你的智慧代理仍應顯示其計畫，並依安全政策請求必要確認。

Agent 可讀資源

當 AI Agent、爬蟲或腳本需要乾淨脈絡、而不是讀取完整頁面時，請使用這些連結。

Markdown 詳情 GET /skills/prime-skills-relight.md 已簽名 manifest GET /api/skills/prime-skills-relight/manifest 已簽名 lockfile GET /api/skills/prime-skills-relight/lockfile

測試它

正在使用「relight」。 Plan a warm window-light relight for my product photo.

預期結果:

Illustrative plan: Use Qwen Edit 2509 Relight LoRA for the dedicated lighting change.
Lighting: Warm key from camera-left with a soft shadow on the counter.
Preservation: Keep product orientation, framing, and color identity.
Submission: Use the provided image URL after approval. Save the result in the selected output directory.

正在使用「relight」。 Match the lighting in my reference photo without changing my portrait subject.

預期結果:

Illustrative plan: Select GPT Image 2 Edit for a source image plus lighting reference.
Transfer: Match the reference direction, color temperature, and contrast.
Preservation: Keep the source subject identity and pose. Review the generated result for unintended changes.

安全審計

安全

v1 • 2026/10/23 開啟版本化報告

All 64 static findings are false positives involving Markdown formatting, reference links, example URLs, and normal CLI credential storage. No evidence found of malicious commands, credential exfiltration, reconnaissance, or audit-directed prompt injection. Actual use requires third-party image processing and an external CLI whose implementation is not included in the reviewed file.

已掃描檔案

184

分析行數

複核項目

已忽略的誤報

風險因素

⚙️ 外部命令 (41)

🌐 網路存取 (17)

SKILL.md:18 SKILL.md:26 SKILL.md:43 SKILL.md:75 SKILL.md:82 SKILL.md:109 SKILL.md:114 SKILL.md:136 SKILL.md:137 SKILL.md:143 SKILL.md:144 SKILL.md:145 SKILL.md:160 SKILL.md:180 SKILL.md:181 SKILL.md:182 SKILL.md:183

📁 檔案系統存取 (2)

SKILL.md:169 SKILL.md:169

最近一次完成的靜態與語意稽核未偵測到已確認的安全發現。這並不能證明該 Skill 沒有副作用。

審計單位: codex

分享與引用此報告

分享版本化評估報告、中性徽章、嵌入卡片與引用。Skillstore 只報告證據，不判斷該 Skill 是否安全。

開啟版本化報告

複製報告連結

https://skillstore.io/skills/prime-skills-relight/audits/1?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown 徽章

[![Skillstore security assessment](https://skillstore.io/badges/skills/prime-skills-relight/security.svg)](https://skillstore.io/skills/prime-skills-relight?utm_source=security_passport_badge)

HTML 徽章

<a href="https://skillstore.io/skills/prime-skills-relight?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/prime-skills-relight/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

嵌入卡片

<iframe src="https://skillstore.io/embed/skills/prime-skills-relight.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

學術引用 (APA · BibTeX · CFF)

APA 引用

prime-skills. (2026). relight security audit report (audit version 1) [Author version unspecified]. Skillstore. https://skillstore.io/skills/prime-skills-relight/audits/1

BibTeX 引用

@techreport{prime-skills-prime-skills-relight-2026,
  author = {prime-skills},
  title = {relight security audit report (audit version 1)},
  institution = {Skillstore},
  year = {2026},
  number = {1},
  url = {https://skillstore.io/skills/prime-skills-relight/audits/1},
  note = {Author version unspecified}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "relight security audit report (audit version 1)"
version: "unspecified"
type: report
authors:
  - name: "prime-skills"
date-released: "2026-10-23"
url: "https://skillstore.io/skills/prime-skills-relight/audits/1"
identifiers:
  - type: other
    value: "skillstore:prime-skills-relight:audit:1"
    description: "Skillstore immutable audit report identifier"

比較變體

4 個可安裝變體

每位作者的技能都保留為獨立的可安裝項目。推薦變體依 Skillstore 證據排序。

為何此變體排名第一

Skillstore 評分最高

作者 / 來源 Skillstore 評分證據置信度 Skillstore 使用量已更新

agentspace-so 推薦

agentspace-so-relight

Skillstore 評分 78

證據置信度高

Skillstore 使用量 32

已更新

2026-10-22

prime-skills 目前

prime-skills-relight

Skillstore 評分 77

證據置信度中

Skillstore 使用量 2

已更新

2026-10-24

runcomfy-com

runcomfy-com-relight

Skillstore 評分 70

證據置信度高

Skillstore 使用量 49

已更新

2026-09-09

doany-ai

doany-ai-relight

Skillstore 評分 50

證據置信度高

Skillstore 使用量 47

已更新

2026-09-09

Skillstore 評分

評分依據證據置信度: 中

架構

可維護性

內容

社群

規範符合度

你可以打造什麼

Refresh product catalog lighting

Plan warm window lighting for an existing product photo while requesting preservation of product orientation and color.

Explore portrait lighting moods

Request golden-hour lighting, soft fill, or a rim light while keeping the portrait subject and pose consistent.

Match a lighting reference

Route a source image and lighting reference to a reference-capable edit model with explicit preservation instructions.

試試這些提示

Beginner: Warm portrait lighting

Use my provided portrait URL to plan a golden-hour relight. Preserve identity and pose. Show the plan before submitting.

Intermediate: Product light setup

Relight my provided product image with a warm 3200K key from camera-left and soft cool fill. Preserve product orientation, framing, and colors.

Advanced: Match a reference

Use my source and lighting-reference URLs with GPT Image 2 Edit. Match reference direction, temperature, and contrast. Preserve the source subject and framing.

Advanced: Plan consistent gallery lighting

Plan consistent soft window lighting across my provided product URLs using Nano Banana 2 Edit. Verify endpoint input semantics and shell-safe arguments before submission.

最佳實務

Specify light source, direction, temperature, and intensity, then state which subject details must remain unchanged.
Use only approved image URLs and confirm permission for third-party processing before submitting sensitive assets.
Pass structured inputs with safe shell quoting and review generated results for identity, framing, and color drift.

避免

Expecting subject identity or product colors to remain exact without inspecting the generated result.
Treating the still-image endpoint as a video relighting or image-generation workflow.
Uploading confidential images without approval or pasting API tokens into prompts.

常見問題

What does this skill change?

It guides changes to lighting direction, color temperature, intensity, and mood in an existing still image.

Which model is the default?

The documented default is Qwen Edit 2509 Relight LoRA. Other edit endpoints are suggested for broader changes or reference-based lighting.

Can it match a reference photo?

The skill describes using GPT Image 2 Edit with a source image and a lighting reference.

Will it preserve the subject exactly?

Preservation is requested through prompts, not guaranteed. Inspect identity, pose, framing, and product colors after generation.

Does processing happen locally?

No. The documented workflow uses authenticated RunComfy model APIs and downloads generated results through the CLI.

Can it relight video or create a new image?

Those tasks are outside this skill. It links to separate generation skills and video workflows.

開發者詳細資訊

作者

prime-skills

授權

MIT

Skillstore 修訂

版本提示

作者未聲明版本。

儲存庫

https://github.com/prime-skills/runcomfy-agent-skills/tree/fca19ae084c2ae4667f5c8affbb99da5103d5832/relight

引用

86ead0df2bd0b9fe4e776a0755cf36b63d41be56

維護新鮮度

2026/10/23

使用量

0 次下載 · 0 次瀏覽

檔案結構

📄 SKILL.md