| name | agent-builder |
| description | Design practical AI agents by choosing minimal tools, clean context, and on-demand knowledge. |
Agent Builder Skill
Use this skill when the user wants to design or extend an agent, harness, or tool-using assistant.
Core ideas
- Keep the loop simple.
- Add tools only when the model needs a real new capability.
- Load domain knowledge on demand instead of bloating the system prompt.
- Isolate noisy subtasks with child agents when context gets messy.
Design checklist
- What is the agent trying to accomplish?
- Which 3-5 tools are truly essential?
- What knowledge should be optional instead of always present?
- What context should stay in the main thread, and what should be isolated?
Anti-patterns
- Too many tools at once
- Huge system prompts with unrelated instructions
- Hardcoded workflows that block model reasoning
- No visibility into which tool or skill was used
Rule of thumb
Give the model a small set of clear capabilities, expose optional expertise through skills, and let the harness stay boring.