python-sdk
Warn
Audited by Gen Agent Trust Hub on Jul 17, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: Multiple code examples in 'references/agent-patterns.md' and 'references/tool-builder.md' demonstrate the use of 'eval()' to process expressions for a calculator tool. This is a highly unsafe pattern that can lead to arbitrary code execution if the input (which is generated or passed by the AI agent) is not strictly validated.
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of the 'inferencesh' library and references other standard libraries such as 'requests', 'aiohttp', and 'tqdm'.
- [PROMPT_INJECTION]: The skill documents patterns for building agents with 'web_search', 'code_execution', and file ingestion capabilities. This creates a surface for Indirect Prompt Injection attacks.
- Ingestion points: Data enters the system via the 'web_search' tool, file uploads in 'agent.send_message()', and remote URLs in 'client.run()'.
- Boundary markers: The examples do not demonstrate the use of boundary markers or delimiters to separate untrusted content from instructions.
- Capability inventory: The SDK supports 'code_execution' through internal tools and documentation suggests building tools using 'eval()' (found in 'references/tool-builder.md' and 'references/agent-patterns.md').
- Sanitization: There is no evidence of sanitization or validation of the untrusted data before it is processed by the agent or executed via tools.
Audit Metadata