deepagents-cli

Pass

Audited by Gen Agent Trust Hub on Aug 16, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill documentation defines a workflow where an AI agent reads context from a project-specific memory file (.deepagents/AGENTS.md). This presents an indirect prompt injection surface where a malicious repository could contain a memory file designed to hijack the agent's behavior.
  • Ingestion points: The agent ingests data from .deepagents/AGENTS.md at the project level.
  • Boundary markers: No explicit delimiters or boundary instructions are mentioned to prevent the agent from obeying instructions embedded within the memory file.
  • Capability inventory: The documented tool provides capabilities for shell execution, file system operations, and network access.
  • Sanitization: The documentation does not describe any sanitization or validation of the content within the memory files.
  • [EXTERNAL_DOWNLOADS]: The skill instructs users to install the deepagents-cli package globally via npm. This involves fetching and running external code from the public npm registry.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 16, 2026, 12:31 AM
Security Audit — agent-trust-hub — deepagents-cli