deep-agents-core

Pass

Audited by Gen Agent Trust Hub on Oct 28, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [Filesystem Management Tools]: The framework includes tools for file operations like reading, writing, and editing. The documentation provides guidance on using virtualized backends and human-in-the-loop approvals (via the interrupt_on parameter) to ensure these capabilities are used securely within a controlled environment.- [Modular Instruction Loading]: The skill defines a system for loading additional capabilities (skills) from directories using a structured SKILL.md format. This design pattern allows for on-demand loading of specialized instructions while maintaining oversight over the agent's available toolset.- [Input Ingestion Surface]: As a developer-facing agent framework, the system is designed to process external data and user input. The provided examples illustrate best practices for managing this data safely, including the use of persistent thread-based memory and explicit system prompts.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 28, 2026, 05:18 AM