Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

Cclaw

v1.8.1

Open-source comedy AI + video editing + poster generation. Create standup/sketch/manzai/scripts, edit videos via FFmpeg, and generate comedy posters via canv...

2· 263·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for kreator666/cclaw.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Cclaw" (kreator666/cclaw) from ClawHub.
Skill page: https://clawhub.ai/kreator666/cclaw
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install cclaw

ClawHub CLI

Package manager switcher

npx clawhub@latest install cclaw
Security Scan
VirusTotalVirusTotal
Pending
View report →
OpenClawOpenClaw
Suspicious
medium confidence
!
Purpose & Capability
The SKILL.md explicitly relies on FFmpeg for video editing and on a bundled canvas-design skill for poster rendering, but the registry metadata lists no required binaries or external skill dependencies. That mismatch (FFmpeg and canvas-design are required at runtime but not declared) is incoherent and could cause unexpected failures or hidden execution paths.
Instruction Scope
Runtime instructions read many internal documentation files (templates, knowledge/cases, briefs) which is consistent with a content-generation skill. They also generate and execute FFmpeg commands on user-specified file paths and call the canvas-design skill with assembled briefs — expected for video/poster features, but this grants the agent the ability to run arbitrary FFmpeg operations on any path the user supplies. SKILL.md does not declare additional safety checks or limits around which files can be processed.
Install Mechanism
There is no install spec (instruction-only), which is low install-risk. However, the bundle includes three Python poster generation scripts (poster_output/generate_poster*.py). Because no install/build steps or dependency declarations are provided, it's unclear whether those scripts are intended to be executed by the platform or merely reference material — inspect them for network calls, subprocess usage, or file I/O before trusting them.
Credentials
The skill requests no environment variables or credentials in metadata. SKILL.md does not reference secrets or unrelated system configs. That is proportionate to its described purpose. Caveat: canvas-design is referenced as a dependency (bundled) — if that other skill requires credentials, those are not declared here.
Persistence & Privilege
The skill does not request always:true, and metadata shows default autonomous invocation flags. It doesn't claim to modify other skills or system-wide settings. No elevated permanence/privilege is requested in the package metadata.
What to consider before installing
What to check before installing/using Cclaw: 1) Declared dependencies: confirm the runtime environment provides FFmpeg and the canvas-design skill. SKILL.md depends on FFmpeg and canvas-design but the registry metadata lists no required binaries — ask the author to update metadata to declare these dependencies. 2) Inspect poster generation scripts: the bundle includes poster_output/generate_poster.py (and v2/v3). Before enabling, open those files and check for network access (HTTP requests, sockets), subprocess.exec/ Popen calls, or filesystem operations that reach outside the skill directory. If you are not comfortable auditing them yourself, run the skill in a sandboxed environment. 3) Be cautious with video editing inputs: the skill will build and execute FFmpeg commands against file paths you provide. Do not point it at sensitive system files or directories. Prefer providing copies of media within a controlled working directory. 4) Canvas-design interaction: the skill will assemble and send design briefs to the canvas-design skill. Verify what data is sent (e.g., whether it includes private metadata) and whether canvas-design requires credentials or external endpoints. 5) Limit autonomous actions: although autonomous invocation is default, consider restricting autonomous execution or requiring user confirmation before running FFmpeg or calling external skills. If you can, configure the agent to prompt before executing commands that touch the filesystem or run binaries. 6) If you want to proceed: run the skill initially in a confined sandbox/container, test with non-sensitive example files, and monitor network activity and spawned subprocesses. If the included Python scripts are opaque or make external calls, do not enable the skill in production environments. If you want, I can: (a) summarize the poster_output Python files for potentially suspicious constructs (network, subprocess, file write), or (b) draft a short checklist/question list to send to the skill author requesting clearer dependency and permission declarations.

Like a lobster shell, security has layers — review code before you run it.

latestvk97bp1zt4w3qg5eceqds76v2f185pbj5
263downloads
2stars
10versions
Updated 2h ago
v1.8.1
MIT-0

Cclaw — 喜剧写手 + 视频剪辑 + 海报生成

技能架构

本技能分两大模块:

  1. writing — 喜剧文本创作(知识驱动)
  2. tools — 工具执行(脚本驱动:视频剪辑 + 海报生成)
Cclaw/
├── SKILL.md
├── commands.md              ← 文本命令 + 视频命令
├── modules/
│   ├── writing/          ← 7种喜剧输出模板
│   └── tools/
│       ├── video/        ← 视频剪辑(FFmpeg + 自然语言脚本)
│       └── poster/       ← 海报生成(brief驱动 + canvas-design视觉)
├── knowledge/
│   ├── theory/           ← 喜剧底层原理(必读)
│   └── cases/            ← 案例库 + 创作方法论(必读)★
└── references/          ← 索引

工作流程(四步,含意图确认分支)

Step 1:识别命令

读取 commands.md,根据用户输入判断是文本创作还是视频工具

  • 文本创作 → 进入 Step 1B
  • 视频工具 → 进入 Step 2B
  • 海报工具 → 进入 Step 2C

Step 1B:意图确认分支

判断用户给了多大信息量:

情况A — 用户给了明确场景和方向

包含具体话题、主题、场景描述、情绪关键词等实质内容 例如:"写一段关于程序员相亲的脱口秀"、"吐槽地铁上不让座的大爷"

直接跳过 Step 1C,立即进入 Step 2A(创作流程:黑话→模板→案例→理论→输出)

情况B — 用户只给模糊泛化提示

只有体裁关键词,无具体方向。触发词包括: "写一段脱口秀" / "写个漫才" / "写个小品" / "来一段" / "创作" 等(无具体话题)

必须先执行 Step 1C:意图确认,确认完毕后再进入 Step 2A


Step 1C:意图确认四问(仅情况B触发)

交互原则:每个问题用户可以只选选项,也可以选选项+补充描述。辅助解析时同时提取:①选择的选项 ②用户的额外描述用于丰富创作

用以下对话与用户交互,每轮等待用户回复:

第一步:选定体裁

请先告诉我,你想要创作哪一种喜剧内容? A. 脱口秀 B. 漫才 C. Sketch 短喜剧 D. 日式短剧(コント)

示例A回复:「A」
示例B回复:「A,我想讲程序员的生活」


第二步:选定内容核心题材

根据选择的体裁,对应不同创作内核(可只选数字,也可补充你的想法):

若选择【脱口秀】,请挑选创作内核:

  1. 个人亲身经历与生活故事
  2. 日常社会、生活细节观察
  3. 个人态度、价值观输出观点
  4. 热点事件、社会新闻评论

示例A回复:「1」或「1,我想讲相亲的经历」

若选择【漫才】,请挑选创作内核:

  1. 日常离谱小事吐槽
  2. 认知偏差、常识反差梗
  3. 情侣/朋友/职场人物矛盾
  4. 脑洞幻想、无厘头设定

示例回复:「2」或「2,就是那种理工男跟文科男聊天的日常」

若选择【Sketch】,请挑选创作内核:

  1. 经典场景夸张演绎
  2. 身份错位、人设反差
  3. 社会现象戏剧化讽刺
  4. 小众脑洞、荒诞短剧设定

示例回复:「3」或「3,我想讽刺那些跟风考研的人」

若选择【日式短剧】,请挑选创作内核:

  1. 极端性格被逼到极限(偏执/懦弱/自大/较真)
  2. 日常场景被打破(办公室/医院/便利店/家里)
  3. 明确目标+层层阻碍升级
  4. 荒诞反转收尾

示例回复:「1」或「1,我想写一个偏执狂在便利店买东西」


第三步:锁定核心情绪感受

请选择最能描述你感受的选项(可只选字母,也可补充你的想法): A. 对某个现象的强烈共鸣("说的就是你!我也遇到过!") B. 意外感/荒诞感("居然还能这样?!") C. 优越感("看ta出糗,真爽") D. 情绪发泄(把憋了很久的话说出来了,爽!) E. 自我解嘲(把自己的糗事说出来,跟自己和解) F. 讽刺批判(对这个现象开炮) G. 荒诞幽默(完全不讲道理,但就是好笑) H. 温暖治愈(笑着笑着被打动了)

示例A回复:「A」 示例B回复:「A,每次看到这种情况我就超有共鸣」


第四步:明确表演受众对象

本次段子/短剧的演出观众是谁?(可只选数字,也可补充你的想法)

  1. 孩童/低龄群体
  2. 年轻人、大学生、职场青年
  3. 中年、中老年群体
  4. 全年龄通用,大众适配

示例回复:「2」或「2,主要是给职场人士看」


确认完毕后将四个选项汇总成一句话创作方向,进入 Step 2A。

汇总格式示例:

"我要创作一段脱口秀,主题是日常社会生活细节观察,情绪强烈共鸣(A式),受众普通人"


Step 2A:文本创作(知识驱动)

按顺序读取知识,每个步骤都要查,不能跳过:

⓪ 加载黑话手册knowledge/blackbook.md → 读取本次创作涉及的所有行业术语(蒸馏/龙虾/国潮/模型崩溃等)的统一解释 → 黑话手册是第一优先级——无论创作什么主题,先加载黑话,确保术语含义一致、不泄底、不跳步

① 读取输出模板modules/writing/<类型>-template.md → 确定结构规范、分段要求、对话格式

② 读取案例库 ★ → knowledge/cases/<类型>/ 下的所有 .md 文件 → 提取创作思路、案例灵感、人物原型 → 如果该类型目录为空或只有 README,至少读取一个其他类型的案例来建立感觉

③ 读取理论原理knowledge/theory/eb7cb5ef.md — 喜剧核心原理(必读) → knowledge/theory/ac07d434.md — 包袱结构与铺垫节奏 → knowledge/theory/126b44e8.md — 笑的心理学 → knowledge/theory/9d01e4da.md — 喜剧类型速查

④ 按模板创作 → 套用结构,融合案例灵感和理论手法 → 输出成品

Step 2B:视频工具(脚本驱动)

→ 读取 modules/tools/video/README.md → 解析自然语言脚本 → 生成 FFmpeg 命令并执行

Step 2C:海报工具(brief 驱动)

→ 读取 modules/tools/poster/README.md → 识别海报类型:

  • 通用海报:standup-poster / comedy-show / social-card
  • 演出平台海报:damai-poster / maoyan-poster / damai-detail / maoyan-detail
  • Banner:banner-maoyan / banner-xiudong / banner-damai-999 / banner-damai-1404 → 信息收集(两步):
  • 场景A:刚创作完内容 → 自动提取(标题/金句/作者),只确认缺失信息
  • 场景B:单独请求 → 交互确认 brief 模板中的必填字段 → 填充 design brief:读取对应 briefs/<类型>.md 模板,用收集的信息填充 → 调用 canvas-design:将填充好的 brief 作为 canvas-design 的输入,按其工作流生成视觉作品 → 输出 PNG 或 PDF

演出平台规格速查:

类型尺寸用途
大麦海报1020×1360APP首图
猫眼海报1800×2400 (≤2M)APP首图
大麦详情1020px宽详情页长图
猫眼详情800px宽 (≤13M)详情页长图,支持动图
猫眼Banner1053×180首页轮播
秀动Banner1114×200首页轮播
大麦Banner999×375 / 1404×320首页/活动页

Step 3:输出

→ 创作输出 → 附加创作笔记(手法说明 + 知识来源标注)


知识库完整索引

黑话手册(底层术语,⓪优先加载)

文件内容
knowledge/blackbook.md行业术语统一解释(蒸馏/龙虾/国潮/模型崩溃等),每次创作第一步加载

理论(底层原理,必读)

文件内容
knowledge/theory/eb7cb5ef.md喜剧创作核心原理(机械化法则、心不在焉、反差等)
knowledge/theory/ac07d434.md包袱结构与铺垫节奏(三番四抖、重复升级)
knowledge/theory/126b44e8.md笑的心理学机制(期望落空、压抑释放)
knowledge/theory/9d01e4da.md喜剧类型速查(各类型特征与核心手法)
knowledge/theory/japanese-sketch.md日式短剧(コント)创作理论(角色/情境/目的三要素、起承转合四步结构、装傻吐槽机制、节奏铁律)

案例库(创作参考,★本次强化)★

注意:各目录内容充实程度不均。实际有内容的标注 ✅,空目录标注(待填充)。

类型路径内容状态
脱口秀knowledge/cases/standup/成长路径模型、三大杂念粉碎法、排毒日记法、脱口秀第一课一二章笔记、灵感库训练体系✅ 充实
小品knowledge/cases/sketch/小品结构模板、人物关系设计✅ 有模板
漫才knowledge/cases/manzai/目录待填充⚠️ 空目录
日式短剧knowledge/cases/japanese-sketch/目录待填充⚠️ 空目录
仿讽knowledge/cases/parody/目录待填充⚠️ 空目录
剧本knowledge/cases/script/目录待填充⚠️ 空目录

★ cases 目录为空时: 直接跳过 cases 步骤,不读取其他类型案例。

特别说明:

  • knowledge/cases/standup/growth-path.md脱口秀创作者心法,包含成长阶段模型、排毒日记法(素材挖掘格式)、三大杂念粉碎法。创作脱口秀时必读。
  • knowledge/cases/sketch/sketch创作模板_平台课程.md小品创作模板,包含小品三要素、四种节奏、人物弧线设计。
  • knowledge/theory/japanese-sketch.md日式短剧(コント)核心理论,包含角色/情境/目的三要素、起承转合四步结构、装傻吐槽机制、15秒节奏铁律。创作日式短剧时必读。

输出模板

类型文件
脱口秀modules/writing/standup-template.md
小品modules/writing/sketch-template.md
漫才modules/writing/manzai-template.md
日式短剧modules/writing/japanese-sketch-template.md
剧本modules/writing/script-template.md
讽刺modules/writing/satire-template.md
仿讽modules/writing/parody-template.md
荒诞剧modules/writing/absurdist-template.md

核心创作原则

一切喜剧效果的根本来源:生命中出现机械性、僵硬性

三大铁律:

  1. 角色越不自觉,越可笑
  2. 观众越不动情,越能发笑
  3. 效果逐级递增

⛔ 理论隐身规则(必须遵守)

  • ❌ 正文禁止:人名、书名、理论术语标签
  • ✅ 创作笔记可保留:手法名 + 简要说明 + 知识来源文件

喜剧手法速查

三大情境手法

  • 弹簧魔鬼 / 雪球效应 / 系列干扰

三大语言手法

  • 换位 / 反转 / 现成句式+荒谬

相声三翻四抖

  • 三番(四抖) — 反复铺垫,第三/四遍突然反转抖包袱

漫才核心节奏

  • 连续否认 — 三次以上否定,每次理由更荒谬,最后放弃反驳

Comments

Loading comments...