role-creator

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the agent-team CLI to generate role packages, passing user-provided strings for descriptions, goals, and scopes as command-line arguments. This presents a surface for command injection if inputs contain shell metacharacters.\n- [EXTERNAL_DOWNLOADS]: It enables the discovery and installation of external skill packages from third-party repositories, such as antfu/skills, which are not included in the trusted vendor list.\n- [REMOTE_CODE_EXECUTION]: The skill is designed to install and subsequently execute code from external sources to complete tasks, which constitutes the execution of unverified third-party code.\n- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface through the ingestion of untrusted user data for role definitions.\n
  • Ingestion points: User input for role description, system-goal, and in-scope/out-of-scope boundaries.\n
  • Boundary markers: No delimiters or safety instructions are used to wrap these inputs in the command templates.\n
  • Capability inventory: Capabilities include local shell command execution and dynamic package installation/execution.\n
  • Sanitization: There is no evidence of input validation or shell-escaping for the fields approved during the brainstorming process.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 11:48 AM