← 返回 Skills 市场
Picasso TikTok
作者
Paul de Lavallaz
· GitHub ↗
· v1.2.0
· MIT-0
304
总下载
0
收藏
1
当前安装
1
版本数
在 OpenClaw 中安装
/install picasso-tiktok
功能描述
Full TikTok/Reels video pipeline: script → TTS voiceover (ElevenLabs) → HeyGen talking avatar → auto-subtitles (Whisper) → ffmpeg compose → 1080x1920 final v...
安全使用建议
Summary of what to consider before installing:
- Metadata mismatch: the registry says no env vars, but SKILL.md requires many API keys. Ask the publisher to correct the metadata and explain why each credential is needed.
- Sensitive keys: supply only throwaway/test API keys or scoped tokens (least privilege). Do not provide high-value production keys (OpenAI, Replicate, ElevenLabs, HeyGen) until you trust the author.
- Third‑party uploads: the pipeline uploads audio to uguu.se — this exposes your audio to an external host. If that is unacceptable, modify the workflow to use your own storage endpoint.
- Filesystem writes & absolute paths: the guide writes cache to /home/ubuntu/... and reads ~/.openclaw; run the skill in an isolated container or dedicated VM and inspect files it creates.
- Undeclared env vars in examples (Cartesia, CARTESIA_VOICE_ID) mean the instructions may call additional services; request a complete list of env vars and endpoints.
- Human approvals are required by the SKILL.md — maintain that manual review step and do not permit fully autonomous runs until you vet behavior.
Recommended actions: run in a sandbox, review and edit SKILL.md to remove unwanted uploads/paths, ask author for corrected metadata and a minimal set of required credentials, and test with non-production API keys.
功能分析
Type: OpenClaw Skill
Name: picasso-tiktok
Version: 1.2.0
The skill bundle automates a complex TikTok video production pipeline using multiple external APIs (ElevenLabs, HeyGen, OpenAI, Replicate) and system tools like ffmpeg and yt-dlp. It exhibits high-risk behaviors including arbitrary shell command execution for video processing and downloading content from external URLs. Notably, the code in SKILL.md includes logic to upload generated audio files to 'uguu.se', an anonymous third-party file-sharing service, to facilitate HeyGen integration. While these capabilities are plausibly required for the stated purpose, the combination of broad shell access and the exfiltration of media to an unauthenticated public host warrants a suspicious classification.
能力评估
Purpose & Capability
The skill's declared registry metadata lists no required env vars or config paths, but SKILL.md explicitly requires ELEVENLABS_API_KEY, ELEVENLABS_VOICE_ID, HEYGEN_API_KEY, YOUR_HEYGEN_AVATAR_ID, OPENAI_API_KEY, REPLICATE_API_TOKEN and system tools (ffmpeg, yt-dlp, Python). Requiring ElevenLabs/HeyGen keys is coherent with TTS/avatar generation, and OpenAI/Replicate could be plausible for Whisper/Runway integration — but the metadata vs instructions mismatch is an incoherence and a poor signal about the publisher's care. The presence of additional service mentions (Cartesia backup, uguu.se upload) increases required capabilities beyond what's stated.
Instruction Scope
Instructions instruct the agent to download videos (gdown, yt-dlp), read Telegram inbound files at ~/.openclaw/media/inbound/, write cache to an absolute path (/home/ubuntu/clawd/...), upload audio to a public file host (uguu.se), call multiple external APIs, and require manual human approval at several steps. Uploading intermediate audio to a third-party file host and writing to absolute filesystem locations are data‑exfiltration and environment assumptions the registry did not disclose. The SKILL.md also references additional env vars (e.g., CARTESIA_API_KEY) not listed in the top 'Required env vars' header, indicating inconsistency between instructions and declared requirements.
Install Mechanism
There is no install spec and no code files — this is instruction-only, which reduces risk from unexpected installers or archived downloads. However, being instruction-only means the agent will perform network calls and local file writes as described; absence of an install spec does not eliminate the concerns in instruction scope and environment access.
Credentials
The SKILL.md expects multiple credentials (ElevenLabs, HeyGen, OpenAI, Replicate) and also references other keys (Cartesia) inside examples. The registry metadata lists none — so the skill is asking for broad, sensitive credentials that were not declared. These credentials are plausible for the stated pipeline, but the mismatch, the number of credentials, and the presence of additional undocumented env vars are disproportionate and should be justified by the publisher.
Persistence & Privilege
The skill does not request always:true and does not declare privileged modifications to other skills. However, the instructions assume write access to filesystem locations (explicit cache path under /home/ubuntu and references to ~/.openclaw), which presumes filesystem permissions and persistence of intermediate files. This is not inherently malicious but should be considered when deciding where to run the skill (use isolated environment).
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install picasso-tiktok - 安装完成后，直接呼叫该 Skill 的名称或使用
/picasso-tiktok触发 - 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
v1.2.0
v1.2: ElevenLabs 3-variation flow, Runway Gen-4.5 B-roll, image-to-video con Nano Banana Pro + animación con Gen-4.5, layouts avanzados, correcciones de Whisper para voseo rioplatense, checklist completo
元数据
常见问题
Picasso TikTok 是什么？
Full TikTok/Reels video pipeline: script → TTS voiceover (ElevenLabs) → HeyGen talking avatar → auto-subtitles (Whisper) → ffmpeg compose → 1080x1920 final v... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 304 次。
如何安装 Picasso TikTok？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install picasso-tiktok」即可一键安装，无需额外配置。
Picasso TikTok 是免费的吗？
是的，Picasso TikTok 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Picasso TikTok 支持哪些平台？
Picasso TikTok 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Picasso TikTok？
由 Paul de Lavallaz（@pauldelavallaz）开发并维护，当前版本 v1.2.0。
推荐 Skills