devflow-automation

Pass

Audited by Gen Agent Trust Hub on Sep 6, 2026

Risk Level: SAFECREDENTIALS_UNSAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • Hardcoded Credentials in Examples: The C# implementation example in the skill includes hardcoded credentials, specifically a password string (password123). While this is presented as a shortcut for developer testing, using hardcoded secrets in source code is a security consideration that should be avoided in production environments.
  • Authentication Bypass Capability: The skill explicitly instructs on "Authentication Bypass" patterns using maui_invoke_action. This allows the agent to reach deep application states by skipping standard security checkpoints like login screens, which is a capability that requires careful oversight.
  • Dynamic Action Definition and Execution: The workflow leverages C# Hot Reload to allow the agent to add new [DevFlowAction] methods at runtime and immediately execute them. This provides a significant surface for dynamic code execution within the target application's process.
  • Indirect Prompt Injection Surface: The agent relies on listing actions (maui_list_actions) and reading their descriptions to decide on its next steps. If an application contains metadata or descriptions controlled by untrusted third parties, it could potentially influence the agent's actions through this discovery process.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 6, 2026, 10:33 AM
Security Audit — agent-trust-hub — devflow-automation