技能 ai-video-generation

📦

ai-video-generation

Name: ai-video-generation
Author: genmedia-labs

內容修訂 r1 中風險 ⚙️ 外部命令🌐 網路存取📁 檔案系統存取

Generate AI Videos with RunComfy

Choosing and prompting video models requires detailed model knowledge. This skill routes each request and prepares focused RunComfy CLI invocations.

支援: Claude Codex Code(CC)

⚠️ 50 差

Maintain this Skill? Claim it ⚠️ 50 差

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-video-generation" from https://skillstore.io/skills/genmedia-labs-ai-video-generation.md and its manifest at https://skillstore.io/api/skills/genmedia-labs-ai-video-generation/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/genmedia-labs-ai-video-generation.md 已簽名 manifest GET /api/skills/genmedia-labs-ai-video-generation/manifest 已簽名 lockfile GET /api/skills/genmedia-labs-ai-video-generation/lockfile

測試它

正在使用「ai-video-generation」。 Create an eight-second widescreen clip of a red kite crossing a windy beach with natural ambient audio.

預期結果:

Selected model: HappyHorse 1.0 for general text-to-video with synchronized audio.
Prompt: A red kite tumbles across a windy beach at golden hour while children chase it. Audio includes wind, gulls, surf, and distant laughter.
Output destination: ./out

正在使用「ai-video-generation」。 Animate my portrait into a six-second vertical clip with a slow head turn and stable camera.

預期結果:

Selected route: image-to-video.
Motion direction: The subject slowly turns toward the camera and smiles while the camera remains still.
Format: six seconds, vertical 9:16.

正在使用「ai-video-generation」。 Create a realistic product rotation and keep the background motionless.

預期結果:

Selected model: Veo 3-1 for controlled physical motion and object permanence.
Direction: The product rotates 180 degrees on its surface. The camera and all background elements remain still.

安全審計

中風險

v1 • 2026/8/13 開啟版本化報告

The skill intentionally installs and invokes the RunComfy CLI, sends prompts and referenced media to a third-party API, and downloads generated files. Most static alerts are false positives caused by Markdown backticks, visible documentation links, and model-selection prose; token storage is documented with restrictive permissions.

已掃描檔案

425

分析行數

複核項目

已忽略的誤報

已確認的安全問題 (1)

中

SKILL.md:199-213 SKILL.md:402-412

Third-Party Processing of Prompts and Media

The workflow sends prompts and user-selected media URLs to RunComfy model endpoints, where external services process the content. Users must avoid confidential or unauthorized media.

The documented audio_url example and the How it works section explicitly describe API submission, polling, and remote result retrieval.

能力複核項目 (17)

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。

高

SKILL.md:407

Hidden file in home directory

- **Token storage**: `runcomfy login` writes the API token to `~/.config/runcomfy/token.json` with m

The skill documents that runcomfy login stores an API token at ~/.config/runcomfy/token.json. This is intentional credential-file access, mitigated by the stated mode 0600 and environment-variable alternative.

中

SKILL.md:32-43

Ruby/shell backtick execution

```bash

This fenced setup example instructs users to install the RunComfy CLI, authenticate, and invoke a remote model. These are intentional external commands, although their purpose is consistent with the skill.

中

SKILL.md:49-51

Ruby/shell backtick execution

```bash

This fenced example instructs users to run npx to install the community skill globally. It is a real external package-manager command with network and supply-chain exposure.

中

SKILL.md:174-183

Ruby/shell backtick execution

```bash