python-sdk
Pass
Audited by Gen Agent Trust Hub on May 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The documentation references the installation of the 'inferencesh' Python package from the official registry and suggests adding related functionality via 'npx' commands. These resources originate from the vendor's ecosystem.
- [COMMAND_EXECUTION]: The skill provides examples for using the SDK to build agents with internal code execution capabilities and describes how to handle tool calls within a host environment.
- [PROMPT_INJECTION]: The skill provides the framework for building AI agents that ingest untrusted user data, which constitutes an indirect prompt injection surface.
- Ingestion points: Untrusted data enters the agent context through 'agent.send_message()' and file upload APIs described in 'SKILL.md' and 'references/agent-patterns.md'.
- Boundary markers: Provided code examples do not explicitly demonstrate the use of boundary markers or system instructions to ignore embedded commands in user-provided content.
- Capability inventory: The SDK allows agents to be configured with significant capabilities, including runtime code execution, web searching, and external webhook integrations.
- Sanitization: The examples focus on programmatic usage and do not include specific patterns for input validation or the sanitization of data returned from external tools.
- [SAFE]: No indicators of obfuscation, hardcoded credentials, malicious persistence, or unauthorized data exfiltration were found within the instructions or reference materials.
Audit Metadata