creating-kiro-agents

Creating specialized AI agents involves generating JSON configurations that define tool restrictions, prompts, and domain workflows. This Skill provides templates for agents with least-privilege access—file operations limited to specific paths, bash commands to an allowlist—enabling you to deploy focused assistants like backend developers or code reviewers with security guardrails built in.

Role-specific agents differ in their tool configurations and prompts. A backend agent gets access to API directories and Express.js patterns; a frontend agent accesses component files and React conventions. This Skill generates domain-specific patterns for each role, applying the same security model but tailoring workflows to the specialist's scope.

Yes. Agent configurations include MCP server settings, per-tool restrictions (fs_read and fs_write with allowedPaths, execute_bash with allowedCommands), prompts, lifecycle hooks, and resources. This Skill generates the complete JSON structure needed to set up these constraints and enable specialized workflows.

No. This Skill generates ready-to-use JSON agent configurations from your requirements—you specify the role, domain, and access scope, and the Skill produces the structured configuration. You don't need to hand-write the JSON or understand the full schema yourself.

Restricted file paths and command allowlists improve security but require careful scoping. Overly tight restrictions may prevent legitimate workflow tasks; too loose defeats the security goal. This Skill applies least-privilege best practices, but you should validate that path and command restrictions match your agent's actual workflow needs.

This Skill provides domain-specific patterns for roles including test-writer, code-reviewer, backend-dev, and frontend-dev. For each use case, it generates an agent configuration with tailored prompts, tool access, and workflow hooks that match the specialist's responsibilities and constraints.