Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
clip-editor
v1.0.0Video clip editing skill for automatically analyzing video content and generating CapCut draft templates. Uses local Whisper for speech transcription, Qwen-V...
⭐ 0· 79·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The skill claims local Whisper + Qwen-VL-Plus + edge-tts for editing and the scripts implement exactly those capabilities. However the registry metadata declares no required binaries (ffmpeg/ffprobe) or env vars even though SKILL.md and scripts explicitly call ffmpeg/ffprobe and reference API keys (QWEN_API_KEY, ANTHROPIC_API_KEY). Omitting those required runtime dependencies/credentials in metadata is an incoherence.
Instruction Scope
Runtime instructions and scripts access local filesystem paths (video file paths and JianYing draft directories), run ffprobe/ffmpeg, extract frames, and construct draft files — all reasonable for video editing. But the instructions/code also send base64-encoded frames and narration prompts to remote services (Qwen-VL endpoint, Anthropic/Claude, edge-tts). That means user video/audio and derived text may be transmitted externally; SKILL.md does not clearly warn or require consent for that transmission.
Install Mechanism
The skill has no declared install spec, but multiple scripts attempt runtime pip installs (openai, anthropic, edge-tts, etc.) using flags like --break-system-packages; these will modify the Python environment without prior declaration. Dynamic installation behavior increases risk and should have been declared in the registry install spec.
Credentials
The registry lists no required env vars, yet code references API keys (QWEN_API_KEY, ANTHROPIC_API_KEY) and will use them if present. Those keys are relevant to the described remote-model features, but not declaring them in metadata is inconsistent and hides the fact that secrets could be used/exfiltrated. No unrelated credentials (AWS, etc.) are requested — so the keys are plausible for the feature, but their absence from metadata is a transparency issue.
Persistence & Privilege
The skill does not request always:true or force installation, and it does not attempt to modify other skills or global agent configuration. It writes output files and checks local JianYing draft directories (expected for its purpose).
What to consider before installing
Key things to consider before installing or running this skill:
- Network/data exfiltration: The code extracts frames and text and will send them to external APIs (Qwen-VL endpoint, Anthropic/Claude, edge-tts). If you will process private or copyrighted video, assume those fragments may be transmitted to third parties unless you review and disable those calls.
- Undeclared requirements: The skill uses ffmpeg/ffprobe and will attempt to pip-install packages at runtime. The registry metadata does not declare these dependencies or install steps. Running it may modify your Python environment (the scripts use --break-system-packages).
- Credentials: The scripts look for QWEN_API_KEY and ANTHROPIC_API_KEY (and will use them if present). The package did not declare these env vars. Only provide API keys if you trust the skill and understand which remote services will receive your data.
- Mitigations: Run the skill in a sandboxed environment or VM, inspect the code locally, or run with network disabled if you only want local Whisper/edge-tts. Prefer creating an isolated Python virtualenv before running so the automatic pip installs cannot alter your system Python. Ask the publisher to update the registry metadata to declare required binaries, explicit optional env vars, and a proper install spec.
- When in doubt: do not supply API keys or private video to this skill until the above inconsistencies are resolved or you run it in an isolated/safe environment.Like a lobster shell, security has layers — review code before you run it.
latestvk971tpna25nrqdnwhfn7fm69j983g8a0
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
