The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the creation of new agent and skill files, which are components that can execute logic. It includes robust mitigations, such as strict input validation using a regular expression (^[a-z0-9]+(-[a-z0-9]+)*$) and explicit checks to reject path traversal characters (.., /, \). This ensures file creation is confined to the intended <appDataDir>\config\plugins\ directory.
[PROMPT_INJECTION]: The skill proactively counters prompt injection by mandating the inclusion of a 'Prompt Defense Baseline' in the system prompt of every subagent it generates. This template instructs the new agents to resist role-play attacks, ignore attempts to override rules, and treat external data as untrusted.
[SAFE]: The workflow follows the principle of least privilege, specifically advising the agent to only grant powerful tools like Bash if explicitly requested by the user and necessary for the task.

agent-creator