agent-builder

Features and purpose

Build, refine, and evolve OpenCode agents with opinionated, research-backed guidance. Combines documentation research (webfetch, exa, context7, grep.app) with architectural best practices. Use for creating new agents, improving existing ones, troubleshooting agent behavior, or analyzing agent design patterns. Provides opinionated recommendations based on current best practices. Examples: - user: "Create an agent for code reviews" → research patterns, design with security focus, recommend tools/permissions - user: "My agent keeps triggering incorrectly" → analyze description, suggest better trigger phrases, improve specificity - user: "Make this agent better at X" → research X domain, suggest prompt improvements, recommend model/temperature - user: "Should I use temperature 0.1 or 0.7 for this agent?" → analyze use case, provide opinionated guidance with rationale - user: "What's the best way to structure this agent's workflow?" → recommend patterns from research, show examples

Quick start

Follow the repository instructions for this skill and your agent. Review the files and dependencies before adding it to your workflow.