video-agent
Warn
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- EXTERNAL_DOWNLOADS (MEDIUM): The setup instructions require
pip install video-agent-claude-skill. This package is hosted on a public registry but authored by a non-trusted source (donghaozhang), which introduces a dependency risk as the underlying code is not verified by this analysis. - COMMAND_EXECUTION (MEDIUM): The skill utilizes a pipeline orchestration feature that executes complex tasks defined in YAML/JSON configurations. This represents dynamic execution logic where the agent's actions are determined at runtime by external file content, potentially leading to unintended command execution if configurations are malicious.
- PROMPT_INJECTION (LOW): The skill has a surface for indirect prompt injection due to how it handles external data.
- Ingestion points: The skill ingests user-provided prompts for media generation, reads configuration from
config.yaml, and accesses local image files for animation. - Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard embedded commands within these inputs.
- Capability inventory: The skill performs network API calls to various AI providers (FAL AI, Google Vertex, ElevenLabs) and possesses file-read/write capabilities via the
video-agentCLI and Python API. - Sanitization: No mention of input validation or prompt sanitization is found in the documentation.
Audit Metadata