free-tts-voice-cloning Security Report — Low Risk | ClawSafe

15 /100

free-tts-voice-cloning

免费文字转语音与克隆 - 完全免费的本地TTS和声音克隆技能

Legitimate local TTS and voice cloning skill with no malicious behavior detected. The skill performs exactly as documented with minor issues (remote script download for Homebrew installation, version unpinned dependencies) that are standard practices for tool installation scripts.

Skill Namefree-tts-voice-cloning

Duration35.4s

Enginepi

✓

Safe to install

This skill is safe to use. The curl|bash pattern in install_dependencies.sh downloads Homebrew from the official GitHub repository, which is a standard and widely-accepted practice. Users should ensure they review installation scripts before execution as a general security practice.

Findings 2 items

Severity	Finding	Location
Low	Unpinned Python dependency version Supply Chain The install_dependencies.sh script installs mlx-audio without version pinning. This could lead to unexpected behavior if the package changes in the future. `python3.10 -m pip install mlx-audio` → Consider pinning the version: pip install mlx-audio==x.x.x to ensure reproducible builds	`install_dependencies.sh:14`
Info	Initial model download from HuggingFace Doc Mismatch SKILL.md states '无需联网' (no internet required) but the model is downloaded from HuggingFace on first run. This is a minor documentation inaccuracy - the skill processes data locally after initial download. `model = load_model('mlx-community/Qwen3-TTS-12Hz-1.7B-Base-8bit')` → Update SKILL.md to clarify: '首次运行需联网下载模型（约3GB），之后完全本地运行'	`voice_cloning_demo.py:47`

Resource	Declared	Inferred	Status	Evidence
Filesystem	`WRITE`	`WRITE`	✓ Aligned	voice_cloning_demo.py writes audio output files via open() with 'wb' mode
Network	`READ`	`READ`	✓ Aligned	load_model() downloads MLX model from HuggingFace on first run - declared in SKI…
Shell	`WRITE`	`WRITE`	✓ Aligned	install_dependencies.sh executes brew install and pip install - declared in SKIL…
Environment	`NONE`	`NONE`	—	No environment variable access found in code
Skill Invoke	`NONE`	`NONE`	—	No inter-skill invocation found
Clipboard	`NONE`	`NONE`	—	No clipboard access found
Browser	`NONE`	`NONE`	—	No browser access found
Database	`NONE`	`NONE`	—	No database access found

3 findings

🔗

Medium External URL 外部 URL

http://127.0.0.1:18789/

README.md:25

🔗

Medium External URL 外部 URL

https://hf-mirror.com

SKILL.md:239

🔗

Medium External URL 外部 URL

https://docs.openclaw.ai/skills/voice-cloning

package.json:19

File Tree

6 files · 19.3 KB · 645 lines

Markdown 2f · 446L Python 1f · 133L JSON 2f · 36L Shell 1f · 30L

├─ 📋 _meta.json JSON 11L · 575 B

├─ 🔧 install_dependencies.sh Shell 30L · 679 B

├─ 📋 package.json JSON 25L · 958 B

├─ 📝 README.md Markdown 119L · 3.0 KB

├─ 📝 SKILL.md Markdown 327L · 10.3 KB

└─ 🐍 voice_cloning_demo.py Python 133L · 3.8 KB

Dependencies 3 items

Package	Version	Source	Known Vulns	Notes
`mlx-audio`	`*`	pip	No	Version not pinned in install script
`ffmpeg`	`latest`	homebrew	No	Installed via Homebrew
`[email protected]`	`latest`	homebrew	No	Standard Python installation

Security Positives

✓ No credential harvesting or API key theft detected

✓ No data exfiltration to external servers during TTS processing

✓ No obfuscated code (base64, eval, etc.) found

✓ No reverse shell or C2 communication patterns

✓ No access to sensitive paths (~/.ssh, ~/.aws, .env)

✓ No hidden functionality different from documentation

✓ Legitimate use of subprocess for tool installation (Homebrew, pip)

✓ Clear and comprehensive legal/ethics disclaimer included

✓ Local-only processing claim aligns with code implementation (MLX framework)

Scan Report

Findings 2 items

File Tree

Dependencies 3 items

Security Positives