agent-creator
Pass
Audited by Gen Agent Trust Hub on Feb 18, 2026
Risk Level: SAFENO_CODEPROMPT_INJECTION
Full Analysis
- [Privilege Escalation] (LOW): The skill provides instructions and templates for using high-privilege configuration options like 'permissionMode: bypassPermissions' and shell-scripted lifecycle hooks. These features allow subagents to execute tools and commands without manual user confirmation.- [Indirect Prompt Injection] (LOW):\n
- Ingestion points: The skill guides the creation of system prompts for subagents that will process user-provided content (SKILL.md).\n
- Boundary markers: The provided templates do not include markers or instructions to delimit or ignore potential injections in the data the subagents process.\n
- Capability inventory: Examples show configurations for 'Bash', 'Edit', and 'hooks' that execute local scripts (SKILL.md).\n
- Sanitization: The skill does not provide patterns for sanitizing or validating the input processed by the generated agents.
Audit Metadata