team-designer

Pass

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill acts as a meta-generator, creating new AI agent instructions based on user input, which constitutes an indirect prompt injection surface.
  • Ingestion points: User-provided requirements, roles, and domain descriptions in Phase 1.
  • Boundary markers: The generated files do not consistently utilize boundary markers or 'ignore' instructions when interpolating user descriptions into the new skill's router and role files.
  • Capability inventory: The skill uses Agent(), Bash(), Write(), and Edit() to create and manage the generated skills.
  • Sanitization: Only basic formatting (kebab-case) is enforced for names; the primary descriptive content is interpolated without significant validation.
  • [COMMAND_EXECUTION]: The skill utilizes the Bash(*) tool to create directory structures and manage files during the scaffolding phase as seen in phases/02-scaffold-generation.md. While these operations are directed at the skill development directory, they represent the execution of shell commands derived from internal logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 8, 2026, 04:09 PM