agent-builder

Claude Agents activate based on their descriptions and assigned tools. Craft a clear, task-focused description that matches user queries, define the specific tools the agent needs (like Read, Grep, or Bash), and select an appropriate model. The agent-builder Skill guides you through defining these discovery-optimized components so it triggers reliably when users ask for that task.

Agent design follows a structured workflow: gather requirements, define the agent's role and scope, specify the tools it needs, choose an appropriate model (Sonnet for complex reasoning, Haiku for speed), and craft a discoverable description. The agent-builder Skill walks you through each phase, from initial design to testing, ensuring your agent handles the task effectively.

Tool assignment depends on what your agent needs to accomplish. Specify exact tools in your agent configuration (e.g., `Read`, `Grep`, `Bash(git diff:*)` for security analysis). Select the model based on task complexity and latency requirements—Sonnet for sophisticated analysis, Haiku for faster, simpler tasks. The agent-builder Skill helps you match both to your use case.

Claude Agents require YAML frontmatter with name, description, tools, model, permissionMode, and skills fields. The agent-builder Skill provides templates and guidance for each field, showing you how to structure them so your agent integrates properly, activates for the right prompts, and delegates tasks correctly.

Yes. Claude Agents support multi-phase workflows and Task tool delegation to subagents. Configure each subagent with its own tools, model, and description, then orchestrate them through your parent agent. The agent-builder Skill covers the full design process, including how to set up subagent coordination and ensure reliable task handoff.

Descriptions that are too vague or generic fail to activate reliably. Avoid ambiguous language; be specific about the agent's scope, the problems it solves, and the tasks it handles. Use domain vocabulary that matches user queries. The agent-builder Skill teaches description optimization techniques so your agent activates for the right user prompts and behaves predictably.