← 返回 Skills 市场
samergmtm22

RoughCut

作者 samerGMTM22 · GitHub ↗ · v1.0.0
darwin ⚠ suspicious
697
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install roughcut-openclaw
功能描述
Run RoughCut headlessly on macOS to generate Final Cut Pro (FCPXML) rough-cut timeline variants from a talking-head video — local-first, no media upload.
安全使用建议
This skill is coherent with its purpose, but exercise normal caution: verify that the RoughCut repo exists at the REPO_ROOT you point to and audit $REPO_ROOT/scripts/openclaw/roughcut.sh before running (it will be executed by the agent). GEMINI_API_KEY is optional — only provide it if you enable fluff removal and you trust the model call. Prefer running the script manually once to confirm behavior and outputs, and ensure OUTPUT_ROOT and REPO_ROOT point to safe directories (so downloaded video and intermediate files aren't saved into sensitive locations).
功能分析
Type: OpenClaw Skill Name: roughcut-openclaw Version: 1.0.0 The skill is classified as suspicious due to its reliance on executing an arbitrary local bash script (`roughcut.sh`) from a configured repository path and its capability to download content from arbitrary user-provided URLs (`--video-url`) using tools like `curl`, as specified in `SKILL.md`. While these capabilities are presented as necessary for the skill's stated purpose of video processing, they introduce significant attack surfaces. The content of `roughcut.sh` is not provided for analysis, and the arbitrary URL download feature could be exploited if the agent is prompted to fetch malicious payloads. There is no clear evidence of intentional malicious behavior (e.g., exfiltration, persistence) within the provided `SKILL.md` itself, but the inherent risks of these powerful operations warrant a 'suspicious' classification.
能力评估
Purpose & Capability
Name/description (headless RoughCut -> FCPXML variants) align with the declared requirements: ffmpeg, python3, node/npm, curl and a shell runner are reasonable for a media-processing runner that invokes local scripts. The skill requires the RoughCut repo to be present locally, which matches the described operation.
Instruction Scope
SKILL.md keeps scope narrow: it asks the agent to verify a local video path or a user-provided download URL and to invoke a script inside the local repo ($REPO_ROOT/scripts/openclaw/roughcut.sh). That is expected, but running a user-local shell script means the agent will execute arbitrary code from the repo — the skill does not bundle or vet that code, so the user must ensure the repo at REPO_ROOT is trusted and matches the referenced GitHub project.
Install Mechanism
No install spec (instruction-only) — nothing is downloaded or installed by the skill itself. This minimizes installation risk; the skill assumes required binaries are already installed.
Credentials
The only declared credential is GEMINI_API_KEY, and the SKILL.md documents it as optional and only used for fluff removal. No unrelated secrets or many environment variables are requested. If fluff removal is disabled, the key is unnecessary.
Persistence & Privilege
always is false and there is no install that persists code or modifies other skill/system configs. The skill does instruct running a local script, but it does not request persistent privileges or force-inclusion.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install roughcut-openclaw
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /roughcut-openclaw 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
Initial release of roughcut-openclaw. - Allows OpenClaw agents to run RoughCut locally on macOS to generate Final Cut Pro (FCPXML) rough-cut timeline variants from talking-head videos. - Supports processing videos from local files or direct HTTPS download URLs without uploading media. - Optionally removes bad takes and "fluff" (off-topic content) using a GEMINI_API_KEY environment variable. - Provides clear user prompts and operational commands, including handling synced folders and download URLs. - Returns output zip file paths for easy import into Final Cut Pro.
元数据
Slug roughcut-openclaw
版本 1.0.0
许可证
累计安装 0
当前安装数 0
历史版本数 1
常见问题

RoughCut 是什么?

Run RoughCut headlessly on macOS to generate Final Cut Pro (FCPXML) rough-cut timeline variants from a talking-head video — local-first, no media upload. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 697 次。

如何安装 RoughCut?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install roughcut-openclaw」即可一键安装,无需额外配置。

RoughCut 是免费的吗?

是的,RoughCut 完全免费(开源免费),可自由下载、安装和使用。

RoughCut 支持哪些平台?

RoughCut 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(darwin)。

谁开发了 RoughCut?

由 samerGMTM22(@samergmtm22)开发并维护,当前版本 v1.0.0。

💬 留言讨论