video-edit
Fail
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [REMOTE_CODE_EXECUTION]: The installation instructions in README.md and SKILL.md suggest piping a remote shell script from the author's GitHub repository directly into the terminal (
curl -sSL https://raw.githubusercontent.com/hoodini/ai-agents-skills/master/install.sh | bash). This is a high-risk execution pattern as it runs unverified remote code. - [COMMAND_EXECUTION]: The skill orchestrates video editing by executing multiple CLI tools, including ffmpeg, ffprobe, and npx hyperframes, via shell commands.
- [DYNAMIC_EXECUTION]: The transcript editor dynamically loads the
@mlc-ai/web-llmlibrary from a remote CDN (https://esm.run/@mlc-ai/web-llm) when the user enables AI suggestions. It also generates and renders HTML compositions at runtime. - [INDIRECT_PROMPT_INJECTION]: The skill ingests transcript data from
transcript.jsonand interpolates it into prompts for AI suggestions and video generation. This represents an indirect prompt injection surface as transcript content could contain malicious instructions. - Ingestion points:
transcript.json(processed byserve_review.pyandapply_review.py). - Boundary markers: None identified in the transcript processing logic.
- Capability inventory: Shell execution of video processing tools (ffmpeg, hyperframes).
- Sanitization: No sanitization of transcript content was observed in the provided scripts.
- [EXTERNAL_DOWNLOADS]: The skill fetches assets and scripts from external domains like
esm.run,cdn.jsdelivr.net, andstatic.heygen.aiduring operation.
Recommendations
- HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/hoodini/ai-agents-skills/master/install.sh - DO NOT USE without thorough review
Audit Metadata