← 返回 Skills 市场
TubeScribe
作者
matusvojtek
· GitHub ↗
· v1.1.8
5202
总下载
7
收藏
8
当前安装
18
版本数
在 OpenClaw 中安装
/install tubescribe
功能描述
YouTube video summarizer with speaker detection, formatted documents, and audio output. Works out of the box with macOS built-in TTS. Optional recommended tools (pandoc, ffmpeg, mlx-audio) enhance quality. Requires internet for YouTube access. No paid APIs or subscriptions. Use when user sends a YouTube URL or asks to summarize/transcribe a YouTube video.
安全使用建议
Install only if you are comfortable with a YouTube URL triggering background processing that fetches captions/comments, writes files locally, generates audio, and may expose transcript/comment text to your active agent or model environment. Review setup.py before accepting optional downloads, avoid private or sensitive videos, and consider changing the workflow to require confirmation before processing a URL.
功能分析
Type: OpenClaw Skill
Name: tubescribe
Version: 1.1.8
This skill bundle is classified as benign. While its functionality involves interactions with the file system, network, and execution of external binaries, which inherently carry risks, the code demonstrates robust security practices. Key indicators include comprehensive URL validation, filename sanitization, safe handling of text passed to subprocess commands (using `json.dumps()` to prevent code injection), and explicit XSS protection in the HTML writer (allowlisting safe tags and stripping event handlers). The `CHANGELOG.md` reveals a history of critical vulnerabilities (code injection, XSS, path traversal, race conditions) that were identified and *fixed*, indicating a proactive approach to security rather than malicious intent. The `SKILL.md` also includes strong prompt injection defenses, instructing the sub-agent not to install software or make unauthorized network calls.
能力评估
Purpose & Capability
The YouTube fetching, comment extraction, document export, and audio generation are coherent with the stated summarizer purpose, but the artifacts also claim processing stays on the user's machine while the workflow has a sub-agent read transcript/comment data to create summaries.
Instruction Scope
SKILL.md tells the agent to spawn a sub-agent immediately when a YouTube URL is sent and run network access, file writes, audio generation, cleanup, and folder opening without an explicit confirmation checkpoint.
Install Mechanism
Setup prompts before optional Homebrew, PyPI, and GitHub-release installs into user-owned tool directories, and the README discloses those downloads; however the runtime no-install wording and model/tool download behavior could be clearer.
Credentials
Network access to YouTube, local output files, temp transcript/comment JSON, and optional TTS/model tooling fit the feature; I found no credential, cookie, login-profile, or unrelated local-data harvesting behavior.
Persistence & Privilege
The skill persists config and queue files under ~/.tubescribe, output under ~/Documents/TubeScribe, optional tools under ~/.openclaw/tools, and temporary source files with cleanup support; there is no evidence of elevated privileges, hidden daemon persistence, or destructive persistence.
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 tubescribe - 安装完成后，直接呼叫该 Skill 的名称或使用
/tubescribe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
v1.1.8
- Description updated: works out of the box, optional tools enhance quality, internet required for YouTube
- Added OpenClaw metadata: declares summarize as required binary
- Added security clarifications: no data uploaded, sub-agent has strict no-install instructions
- Expanded setup.py documentation in README
- Fixed display name on ClawHub: TubeScribe (was Tubescribe)
v1.1.7
TubeScribe 1.1.7
- Updated documentation in SKILL.md and README.md for clarity on local vs. network operations.
- Clarified no paid APIs or subscriptions are required, but network is used to fetch YouTube data.
- Improved instructions regarding the handling of source/output paths and first-time setup.
- Minor corrections to workflow steps and documentation formatting for setup and configuration.
v1.1.6
Fixed long code line in SKILL.md causing horizontal scroll on ClawHub page
v1.1.5
MLX-Audio TTS (3-4x faster on Apple Silicon), 12 bug fixes, 4 security fixes, yt-dlp max_comments correct 5-value format, per-user secure temp directory, SKILL.md temp path fix
v1.1.4
Fix: wrapped long line in SKILL.md that caused horizontal scroll on ClawHub page
v1.1.3
## v1.1.3 — Bug Fixes, Security Hardening & Config Improvements
### Bug Fixes
- Fixed URL handling: normalize /live/, /shorts/ and other YouTube URL formats
- Fixed false-positive error detection (only check error patterns when command actually fails)
- Fixed queue file locking: proper lock-before-open semantics with os.open()+os.fdopen()
- Fixed timezone-aware comparison in stale job detection (was crashing on some systems)
- Fixed link double-encoding in HTML output (URLs with & were breaking)
- Fixed built-in TTS (macOS say) for long summaries: uses temp file to avoid CLI arg length limits
- Fixed headers and table cells now render bold/italic/links correctly in HTML output
- Cleanup now removes TTS temp script files too
### Security
- Zip/tar path traversal validation when installing pandoc/yt-dlp (prevents zip-slip attacks)
- Added single-quote escaping in HTML output
- Added safety rule in SKILL.md to prevent sub-agents from installing software
### Improvements
- Kokoro voice blend and speed now read from config (no more hardcoded values)
- Config uses nested key access throughout (audio.format instead of flat keys)
- Renamed config.set() to config.set_value() to avoid shadowing Python builtin
- setup.py delegates to tubescribe.py for find_ytdlp/find_kokoro (DRY, no duplication)
- Uses find_ytdlp() helper with multi-path search instead of raw subprocess calls
- SKILL.md uses relative paths (portable across installations)
- TTS respects tts_engine config setting (kokoro/builtin/disabled)
- Better error message when macOS say command not found
- Removed unreliable live_stream error pattern
- Default Kokoro path updated to ~/.openclaw/tools/kokoro
v1.1.2
Bug fixes: stale queue crash, config corruption, HTML table/XSS fixes, URL validation. Removed python-docx fallback and afconvert dependency. Simplified builtin TTS (say → WAV direct). Added queue file locking, blockquotes. Improved comment formatting (dash + italic).
v1.1.1
Non-blocking async workflow, improved queue processing, viewer sentiment and best comments
v1.1.0
Step timing, bold headings, improved comment formatting
v1.0.8
Fixed RTF→HTML, added Markdown support, updated config docs
v1.0.7
Simplified frontmatter
v1.0.6
Fix publish
v1.0.5
Re-publish: Fixed RTF→HTML, added Markdown support, updated docs
v1.0.4
Fixed RTF references - replaced with HTML. Added Markdown format support. Updated config documentation.
v1.0.3
Fixed display name capitalization (TubeScribe not Tubescribe); restored Markdown in description
v1.0.2
Fixed description: outputs are DOCX or HTML (not Markdown)
v1.0.1
Cleanup: removed unused RTF writer (HTML is the fallback now)
v1.0.0
Initial release: YouTube video summarizer with speaker detection, documents, and audio
元数据
常见问题
TubeScribe 是什么？
YouTube video summarizer with speaker detection, formatted documents, and audio output. Works out of the box with macOS built-in TTS. Optional recommended tools (pandoc, ffmpeg, mlx-audio) enhance quality. Requires internet for YouTube access. No paid APIs or subscriptions. Use when user sends a YouTube URL or asks to summarize/transcribe a YouTube video. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 5202 次。
如何安装 TubeScribe？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install tubescribe」即可一键安装，无需额外配置。
TubeScribe 是免费的吗？
是的，TubeScribe 完全免费（开源免费），可自由下载、安装和使用。
TubeScribe 支持哪些平台？
TubeScribe 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 TubeScribe？
由 matusvojtek（@matusvojtek）开发并维护，当前版本 v1.1.8。
推荐 Skills