📦
ai-video-generation
74Generate AI Videos with RunComfy
Choosing a video model and formatting its inputs takes time. This skill selects a suitable RunComfy route and prepares the generation request.
設計與創意 查看
技能 ai-image-generation
📦
ai-image-generation
內容修訂 r1 高風險 ⚙️ 外部命令🌐 網路存取📁 檔案系統存取
Generate Images with the Right RunComfy Model
Choosing an image model and its parameters is difficult. This skill routes each request to a suitable RunComfy model and prepares the generation command.
支援: Claude Codex Code(CC)
安裝前請停止並請求確認。
請審查安裝計畫，並在變更檔案前取得使用者明確同意。
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 "ai-image-generation" from https://skillstore.io/skills/skills-collective-ai-image-generation.md and its manifest at https://skillstore.io/api/skills/skills-collective-ai-image-generation/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.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。
測試它
正在使用「ai-image-generation」。 Create a wide photoreal product image of a ceramic mug in warm morning light.
預期結果:
- Selected FLUX 2 Klein 9B for a polished general-purpose image.
- Prepared a wide final-quality generation with explicit subject and lighting details.
- Saved returned images in the selected output folder.
正在使用「ai-image-generation」。 Design a launch poster whose headline reads exactly "AURORA Spring 2026".
預期結果:
- Selected GPT Image 2 for precise headline rendering and layout control.
- Prepared a landscape poster request with the headline quoted exactly.
正在使用「ai-image-generation」。 Keep the person unchanged and replace the background with a rainy city street.
預期結果:
- Selected Nano Banana 2 Edit for identity-preserving image editing.
- Placed preservation requirements before the requested background change.
安全審計
高風險v1 • 2026/7/14 開啟版本化報告
Most alerts are false positives from Markdown syntax, documentation links, URL placeholders, and model comparison text. The skill intentionally runs external packages and RunComfy commands. User-derived shell arguments need robust escaping, and prompts or references leave the local environment.
1
已掃描檔案
497
分析行數
15
複核項目
0
已忽略的誤報
已確認的安全問題 (2)
高
Unsafe shell construction for user-derived prompts
The skill claims prompt content has no shell-injection surface, but examples place JSON inside shell quotes. An unescaped apostrophe can terminate quoting before RunComfy receives the argument.
The examples use a shell command with single-quoted JSON, while line 481 explicitly claims quotes cannot create injection. Shell parsing occurs before the CLI receives its argument.
中
User content is processed by a third-party service
RunComfy receives prompt bodies and reference URLs, polls remote jobs, and returns generated files. Sensitive prompts or signed reference URLs therefore leave the local environment.
The source explicitly states that the CLI posts request bodies and that RunComfy servers fetch references and web results.
能力複核項目 (15)
這些是真實的本機能力，對此技能而言可能是預期行為，因此需要複核，但不會按已確認的惡意行為計分。
中
Ruby/shell backtick execution
```bash
The enclosed setup block directs users to install or execute an npm package, authenticate, and invoke RunComfy. These are intentional external commands with package, network, and credential exposure.
中
Ruby/shell backtick execution
```bash
The enclosed command uses npx to fetch and execute a skill installer with global scope. This is intentional external command execution and carries normal package supply-chain risk.
中
Ruby/shell backtick execution
```bash
The enclosed example executes the RunComfy CLI, sends a prompt to a remote model, and writes downloaded output. This is an intended external command with network and filesystem effects.
中
Ruby/shell backtick execution
```bash
The enclosed example executes the RunComfy CLI, sends a prompt to a remote model, and writes downloaded output. This is an intended external command with network and filesystem effects.
中
Ruby/shell backtick execution
```bash
The enclosed example executes the RunComfy CLI, sends a prompt to a remote model, and writes downloaded output. This is an intended external command with network and filesystem effects.
中
Ruby/shell backtick execution
```bash
The enclosed example executes the RunComfy CLI, sends a prompt to a remote model, and writes downloaded output. This is an intended external command with network and filesystem effects.
中
Ruby/shell backtick execution
```bash
The enclosed example executes the RunComfy CLI, sends a prompt to a remote model, and writes downloaded output. This is an intended external command with network and filesystem effects.
中
Ruby/shell backtick execution
```bash
The enclosed example executes the RunComfy CLI, sends a prompt to a remote model, and writes downloaded output. This is an intended external command with network and filesystem effects.
中
Ruby/shell backtick execution
```bash
The enclosed example executes the RunComfy CLI, sends a prompt to a remote model, and writes downloaded output. This is an intended external command with network and filesystem effects.
中
Ruby/shell backtick execution
```bash
The enclosed edit example executes RunComfy with a remote image reference and writes output locally. This is intentional external command execution with network and filesystem effects.
中
Ruby/shell backtick execution
```bash
The enclosed edit example executes RunComfy with a remote image reference and writes output locally. This is intentional external command execution with network and filesystem effects.
中
Ruby/shell backtick execution
```bash
The enclosed edit example executes RunComfy with a remote image reference and writes output locally. This is intentional external command execution with network and filesystem effects.
中
Ruby/shell backtick execution
The skill classifies the user request into one of the t2i or i2i routes above and invokes `runcomfy
This line states that the skill invokes RunComfy, posts request data, polls a remote service, and downloads results. External command execution is an intended operational capability.
中
Ruby/shell backtick execution
- **Install via verified package manager only.** This skill instructs the operator to install the CL
This line instructs the operator to install or execute the RunComfy npm package. The package-manager source is identified, but executing third-party packages remains a real supply-chain risk.
中
Ruby/shell backtick execution
- **Scope of bash usage**: declared `allowed-tools: Bash(runcomfy *)`. The skill never instructs the
The declared Bash scope authorizes all RunComfy subcommands, confirming intentional external command execution. The scope is limited to one CLI, but that CLI can use network and filesystem resources.
風險因素
⚙️ 外部命令 (50)
SKILL.md:6 SKILL.md:15 SKILL.md:25 SKILL.md:31-45 SKILL.md:45-51 SKILL.md:51-53 SKILL.md:53-61 SKILL.md:61-66 SKILL.md:66-71 SKILL.md:71 SKILL.md:76-81 SKILL.md:81-82 SKILL.md:82-86 SKILL.md:86-91 SKILL.md:91-96 SKILL.md:96-101 SKILL.md:101-106 SKILL.md:106-107 SKILL.md:107-111 SKILL.md:111 SKILL.md:116-117 SKILL.md:117-123 SKILL.md:123-128 SKILL.md:128-131 SKILL.md:131-133 SKILL.md:133-136 SKILL.md:136-138 SKILL.md:138-143 SKILL.md:143-148 SKILL.md:148-153 SKILL.md:153-158 SKILL.md:158-163 SKILL.md:163-168 SKILL.md:168-174 SKILL.md:174 SKILL.md:181-182 SKILL.md:182-183 SKILL.md:183-184 SKILL.md:184-192 SKILL.md:192-201 SKILL.md:201-205 SKILL.md:205-209 SKILL.md:209-216 SKILL.md:216-222 SKILL.md:222-229 SKILL.md:229 SKILL.md:230 SKILL.md:236-243 SKILL.md:243-247 SKILL.md:247-254
🌐 網路存取 (50)
SKILL.md:19 SKILL.md:25 SKILL.md:27 SKILL.md:47 SKILL.md:71 SKILL.md:76 SKILL.md:91 SKILL.md:96 SKILL.md:101 SKILL.md:106 SKILL.md:111 SKILL.md:116 SKILL.md:123 SKILL.md:131 SKILL.md:136 SKILL.md:138 SKILL.md:143 SKILL.md:148 SKILL.md:153 SKILL.md:158 SKILL.md:168 SKILL.md:175 SKILL.md:186 SKILL.md:223 SKILL.md:268 SKILL.md:316 SKILL.md:317 SKILL.md:344 SKILL.md:345 SKILL.md:346 SKILL.md:347 SKILL.md:348 SKILL.md:356 SKILL.md:364 SKILL.md:377 SKILL.md:391 SKILL.md:411 SKILL.md:445 SKILL.md:446 SKILL.md:447 SKILL.md:448 SKILL.md:449 SKILL.md:450 SKILL.md:451 SKILL.md:452 SKILL.md:469 SKILL.md:492 SKILL.md:493 SKILL.md:494 SKILL.md:495
📁 檔案系統存取 (5)
審計單位: claude
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。
複製報告連結
https://skillstore.io/skills/skills-collective-ai-image-generation/audits/1?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_reportMarkdown 徽章
[](https://skillstore.io/skills/skills-collective-ai-image-generation?utm_source=security_passport_badge)HTML 徽章
<a href="https://skillstore.io/skills/skills-collective-ai-image-generation?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/skills-collective-ai-image-generation/security.svg" alt="Skillstore security assessment" loading="lazy"></a>嵌入卡片
<iframe src="https://skillstore.io/embed/skills/skills-collective-ai-image-generation.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
skills-collective. (2026). ai-image-generation security audit report (audit version 1) [Author version unspecified]. Skillstore. https://skillstore.io/skills/skills-collective-ai-image-generation/audits/1BibTeX 引用
@techreport{skills-collective-skills-collective-ai-image-generation-2026,
author = {skills-collective},
title = {ai-image-generation security audit report (audit version 1)},
institution = {Skillstore},
year = {2026},
number = {1},
url = {https://skillstore.io/skills/skills-collective-ai-image-generation/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: "ai-image-generation security audit report (audit version 1)"
version: "unspecified"
type: report
authors:
- name: "skills-collective"
date-released: "2026-07-14"
url: "https://skillstore.io/skills/skills-collective-ai-image-generation/audits/1"
identifiers:
- type: other
value: "skillstore:skills-collective-ai-image-generation:audit:1"
description: "Skillstore immutable audit report identifier"
比較變體
24 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 使用量較高
作者 / 來源 Skillstore 評分 證據置信度 Skillstore 使用量 已更新
tool-belt 推薦
tool-belt-ai-image-generation
Skillstore 評分 78
證據置信度 高
Skillstore 使用量 31
已更新
2026-09-09
runcomfy-com-ai-image-generation
Skillstore 評分 78
證據置信度 高
Skillstore 使用量 29
已更新
2026-09-09
inference-shell-ai-image-generation
Skillstore 評分 78
證據置信度 高
Skillstore 使用量 17
已更新
2026-09-09
infsh-skills-ai-image-generation
Skillstore 評分 69
證據置信度 高
Skillstore 使用量 16
已更新
2026-09-09
inference-skills-ai-image-generation
Skillstore 評分 69
證據置信度 高
Skillstore 使用量 16
已更新
2026-09-09
inference-sh-skills-ai-image-generation
Skillstore 評分 69
證據置信度 高
Skillstore 使用量 14
已更新
2026-09-09
gencraft-labs-ai-image-generation
Skillstore 評分 69
證據置信度 中
Skillstore 使用量 7
已更新
2026-10-11
genmedia-labs-ai-image-generation
Skillstore 評分 69
證據置信度 中
Skillstore 使用量 7
已更新
2026-09-09
skills-shell-ai-image-generation
Skillstore 評分 50
證據置信度 高
Skillstore 使用量 13
已更新
2026-09-09
prime-skills-ai-image-generation
Skillstore 評分 50
證據置信度 中
Skillstore 使用量 7
已更新
2026-09-09
agentspace-so-ai-image-generation
Skillstore 評分 38
證據置信度 高
Skillstore 使用量 19
已更新
2026-10-22
halt-catch-fire-ai-image-generation
Skillstore 評分 38
證據置信度 中
Skillstore 使用量 57
已更新
2026-09-09
inference-sh-ai-image-generation
Skillstore 評分 38
證據置信度 中
Skillstore 使用量 50
已更新
2026-09-09
inference-sh-9-ai-image-generation
Skillstore 評分 38
證據置信度 中
Skillstore 使用量 26
已更新
2026-09-09
inference-sh-8-ai-image-generation
Skillstore 評分 38
證據置信度 中
Skillstore 使用量 16
已更新
2026-09-09
skills-collective-ai-image-generation
Skillstore 評分 38
證據置信度 中
Skillstore 使用量 10
已更新
2026-09-09
danielgwilson-ai-image-generation
Skillstore 評分 38
證據置信度 中
Skillstore 使用量 10
已更新
2026-09-09
Skillstore 評分
評分依據 證據置信度: 中 55
架構
85
可維護性
87
內容
65
社群
83
規範符合度
你可以打造什麼
Create campaign artwork
Select a model for exact headlines, brand layouts, or polished product visuals and prepare the matching request.
Produce rapid concepts
Generate low-resolution variants for storyboards, moodboards, and early creative review before requesting final output.
Automate image workflows
Choose reproducible model parameters and output settings for scripts, content pipelines, or repeated asset generation.
試試這些提示
Generate a simple image
Create an image of [subject] in [setting]. Use a [style] look, [lighting], and a [aspect ratio] composition.
Create a poster with text
Create a [format] poster for [product]. The headline must read exactly "[headline]" in [font style]. Use [colors] and [layout].
Edit a reference image
Edit [reference URL]. Preserve [identity, pose, or layout]. Change only [target element] to [new appearance]. Keep all other details unchanged.
Plan a multi-stage campaign
Create [count] draft concepts for [campaign]. Compare suitable models, generate economical drafts, select one direction, then prepare a final high-resolution version.
最佳實務
- State the subject first, then describe the scene, lighting, style, and output constraints.
- Use economical draft settings for exploration, then increase resolution or quality only for selected finals.
- Provide only approved reference URLs and confirm before enabling web grounding or sending sensitive content.
避免
- Do not request exact typography from a model that is weak at rendering text.
- Do not expose secrets, private signed URLs, or confidential prompts to remote model services.
- Do not interpolate unescaped user text directly into shell-quoted command arguments.
常見問題
Does this skill generate images locally?
No. It sends requests to RunComfy model endpoints and downloads the generated results.
Which model does the skill choose by default?
It prefers FLUX 2 Klein 9B for general text-to-image work and Nano Banana 2 Edit for general editing.
Can it create posters with exact text?
Yes. It routes typography-focused requests to GPT Image 2 and quotes required text in the prompt.
Can it preserve a person during an edit?
It can request identity-preserving edits, but consistency depends on the model, reference quality, and requested change.
What credentials are required?
A RunComfy login or RUNCOMFY_TOKEN is required. The CLI stores interactive credentials in a protected user configuration file.
Can model costs and availability change?
Yes. Check current RunComfy model pages before large batches or production use.
開發者詳細資訊
授權
MIT
Skillstore 修訂
r1
版本提示
作者未聲明版本。
引用
1a7113106876f4a10455028598215006e9a2d31b
維護新鮮度
2026/7/18
使用量
0 次下載 · 0 次瀏覽
檔案結構
📄 SKILL.md
skills-collective 的更多技能
查看全部📦
ai-music
50Create and Edit AI Music with RunComfy
Choosing a music model and valid parameters can slow production. This skill routes each request and prepares a RunComfy command for generation or audio editing.
設計與創意 查看
📦
image-to-video
50Animate Still Images with the Right RunComfy Model
Choosing an image-to-video model can waste time and credits. This skill routes each request to a suitable RunComfy model and provides focused prompting guidance.
設計與創意 查看
📦
video-edit
38Edit Videos with Intent-Routed AI Models
Choosing the wrong video model wastes time and produces inconsistent edits. This skill selects a suitable RunComfy model and builds model-specific instructions.
設計與創意 查看
相關技能
查看全部📦
imagegen
82Create and Edit Project Images
由 openai
Producing consistent raster assets requires precise prompts, careful edits, and reliable output handling. This skill structures requests, creates or edits visuals, validates results, and saves project-ready files.
設計與創意 查看
📦
controlnet-pose
79建立姿勢控制的圖片與影片
由 runcomfy-com
姿勢控制生成通常需要選擇正確的模型和輸入格式。此技能會將圖片和影片請求路由至 RunComfy 工作流程。
設計與創意 查看
📦
image-outpainting
79將影像延伸至畫框之外
由 agentspace-so
裁切過的影像通常需要更寬的畫布，以用於橫幅、社群貼文或產品版面配置。此 skill 會引導 Claude、Codex 和 Claude Code 使用 RunComfy 模型進行影像外補，同時保留原始主體。
設計與創意 查看
📦
relight
78使用 RunComfy 為影像重新打光
由 agentspace-so
變更影像照明通常需要重新拍攝或手動編輯。此技能可協助 Claude、Codex 或 Claude Code 將靜態影像的重新打光請求導向 RunComfy 模型。
設計與創意 查看
📦
flux-kontext
78使用 Flux Kontext Pro 編輯圖片
由 agentspace-so
精準的圖片編輯需要清楚的模型路由與聚焦的提示詞。此技能協助 Claude、Codex 和 Claude Code 透過 RunComfy 呼叫 Flux 1 Kontext Pro，並使用實用的提示詞模式。
設計與創意 查看
📦
native-transparent-imagegen
86Generate Native-Alpha Transparent Images
由 zseven-w
Image generators can return opaque files or painted checkerboards when transparency is required. This skill generates native-alpha assets and validates every untouched output.
設計與創意 查看