Analyzing AgentScope Library
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- EXTERNAL_DOWNLOADS (HIGH): The skill instructs the agent to clone a GitHub repository from
https://github.com/agentscope-ai/agentscopeand install theagentscopePython package. Sinceagentscope-aiis not within the defined Trusted GitHub Organizations, these operations are classified as high risk. - COMMAND_EXECUTION (MEDIUM): The skill executes
python view_agentscope_module.py. This script is local to the skill but its source code is not provided for review, representing the execution of unverified logic. - PROMPT_INJECTION (MEDIUM): This skill exhibits a Category 8 (Indirect Prompt Injection) surface. It ingests untrusted data from an external repository (tutorials and examples) using
lsandcat. This content is then used for 'analysis and decision-making' by the agent. If the external repository contains malicious instructions, they could influence the agent's subsequent actions.
Recommendations
- AI detected serious security threats
Audit Metadata