Skillv1.0.0

ClawScan security

Course Builder Agent · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignApr 30, 2026, 3:16 AM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: The skill's files and instructions are coherent with its stated purpose (automatically building slide+TTS+subtitle videos); nothing indicates intentional misdirection or hidden exfiltration, though there are a few minor omissions to verify before use.
Guidance: This skill appears to do what it claims — build slides, synthesize TTS, create subtitles and compose a video. Before installing or running it: 1) Ensure ffmpeg and ffprobe are installed from your platform package manager (the script calls them but they aren't auto-installed). 2) Review and be comfortable with pip installing edge-tts and Pillow in your environment (use a virtualenv). 3) If you plan to use the optional 数字人/MCP feature, know that the script currently only prints a placeholder; network/upload behavior would require separate configuration and credentials that are not declared here — review any integration you add. 4) The script attempts to import a zh_slides helper from a relative video-craft-pro path if present — verify that dependency or let the fallback image-generation run. 5) Run first on non-sensitive input and inspect outputs to confirm no unexpected network activity. If you need higher assurance, request the author to (a) declare runtime binary requirements (ffmpeg/ffprobe) in the manifest and (b) explicitly document and implement any external MCP uploads and credential usage.

Review Dimensions

Purpose & Capability: okName/description match what the code does: parse input text, generate slides, call edge-tts for speech, build subtitles and use ffmpeg to compose video. The metadata also declares dependency on video-craft-pro and edge-tts which aligns with the behavior.
Instruction Scope: noteSKILL.md and scripts/build.py limit actions to reading the provided input text (and optional photo), generating slides locally, invoking edge-tts, ffprobe/ffmpeg, and composing a local MP4. The docs mention a 飞影数字人 (MCP) interface for digital-human cloning, but the included script does not implement network upload — it only prints a placeholder. No instructions to read unrelated system files or to transmit data to unknown endpoints are present.
Install Mechanism: noteclaw.json includes pip install commands for edge-tts and Pillow (standard PyPI installs) which is reasonable. There are no downloads from untrusted URLs. However ffmpeg/ffprobe (binaries invoked by the script) are required at runtime but are not declared as required binaries in metadata or installed by the install step — this is an operational omission to address.
Credentials: noteNo environment variables or credentials are requested in the manifest or scripts. The SKILL.md references optional third-party services (video-craft-pro and 飞影数字人/MCP); video-craft-pro is declared as a required skill, but the MCP integration would require credentials/configuration that are not declared or implemented here. The lack of declared credentials is consistent with the code as-is, but if you plan to enable MCP you will need to provide credentials elsewhere.
Persistence & Privilege: okThe skill does not request persistent or elevated platform privileges (always:false). It does not modify other skills or global agent settings. It runs as a user-level script and uses temporary directories for intermediate files.