uloop-execute-dynamic-code

Warn

Audited by Gen Agent Trust Hub on May 6, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the terminal to execute C# code via a CLI tool: uloop execute-dynamic-code --code '<code>'. This allows the agent to run arbitrary shell commands with strings it constructs at runtime.
  • [REMOTE_CODE_EXECUTION]: The core functionality of the skill is to write and execute code dynamically. While the instructions list 'Forbidden' namespaces like System.IO.*, these restrictions are likely implemented as simple string-based blacklists at the tool level, which can often be bypassed using reflection or alternative C# namespaces.
  • [PROMPT_INJECTION]: The skill ingests untrusted user input via the $ARGUMENTS placeholder. There are no boundary markers or instructions to treat this input as data rather than instructions, creating a surface for indirect prompt injection where a user could trick the agent into generating malicious C# code that exfiltrates project data or modifies the host system.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 6, 2026, 05:51 AM