knowledge-short-video
Warn
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructs the agent to utilize external tools such as the 'viral-title' skill from 'github.com/kangarooking/kangarooking-skills' and 'npx hyperframes', which involve downloading and executing code from repositories not included in the trusted vendor list.
- [COMMAND_EXECUTION]: The production workflow relies heavily on the execution of various shell commands and Python scripts (e.g., 'resync.py', 'render.py', 'concat-wav.sh', and 'puppeteer' commands) to process media assets and manage task directories.
- [DYNAMIC_EXECUTION]: The skill provides a specific mechanism (Mode C) for integrating custom TTS engines where the agent is instructed to read source code from a user-provided path ('config_client_code_path') and dynamically modify its internal logic to call the provided SDK functions. This presents a high risk of arbitrary code execution if the provided source is malicious.
- [PROMPT_INJECTION]: The skill ingests untrusted user content, such as zero-shot knowledge points or existing documents, to generate video scripts. The lack of explicit boundary markers or sanitization logic when interpolating this content into the video composition template ('composition.html') or passing it to TTS APIs creates a vulnerability surface for indirect prompt injection.
- [SAFE]: The skill demonstrates secure credential management practices by explicitly instructing users to store sensitive API keys in environment variables rather than hardcoding them within configuration files like 'tts-config.json'.
Audit Metadata