Testing Skills With Subagents
Pass
Audited by Gen Agent Trust Hub on May 30, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill employs strong directive language and role-play framing (e.g., "IMPORTANT: This is a real scenario. Choose and act.") to simulate high-pressure environments. This methodology is used to verify that the agent adheres to its primary instructions even when presented with conflicting goals or simulated constraints. These patterns are used for quality assurance of the agent's behavior rather than to bypass safety filters or core instructions.
- [COMMAND_EXECUTION]: The documentation includes examples of standard shell commands (e.g.,
ls,grep) used to browse and search the local skill library. These commands are limited to local filesystem operations within standard platform directories (~/.claude/skills/) and do not involve unauthorized access or network operations.
Audit Metadata