learning-sdk-integration
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- EXTERNAL_DOWNLOADS (LOW): The skill depends on 'agentic-learning-sdk' and 'agentic-learning' from the 'letta-ai' organization. This organization is not included in the pre-approved trusted sources list. Additionally, there is an inconsistency in package naming between the requirements file and the documentation snippets, which could lead to supply chain confusion.
- PROMPT_INJECTION (LOW): The skill is vulnerable to indirect prompt injection through user-supplied messages.
- Ingestion points: The 'message' parameter in the 'UniversalLLMAgent.chat' method located in 'scripts/universal_examples.py', and similar inputs in documentation snippets.
- Boundary markers: Absent. User input is interpolated directly into calls to LLM providers without delimiters or instructions to ignore embedded commands.
- Capability inventory: The skill is capable of performing network operations via LLM provider SDKs (OpenAI, Anthropic, Gemini) and utilizes the 'agentic-learning-sdk' for persistent state management.
- Sanitization: No input sanitization or validation is performed on the user-provided content.
- DATA_EXFILTRATION (LOW): The skill performs network requests to domains associated with LLM providers (api.openai.com, api.anthropic.com, and Google APIs). These domains are not present in the defined whitelist for network operations, though they are necessary for the skill's primary function.
Audit Metadata