agent-development

Agent development converts user requests into validated Markdown files with YAML frontmatter defining trigger conditions, system prompts, model settings, tools, and example blocks. Structure your agent with a clear identifier, when-to-use guidance, second-person system prompt, and 2–4 examples, then validate before deployment.

System prompts use second-person language to define agent responsibilities, processes, and expected outputs. Include specific instructions for how the agent should behave, what tasks it handles, and the format of its responses to ensure consistent and reliable agent execution.

After defining your agent in Markdown with YAML frontmatter, use agent-validation tooling to test trigger conditions, system prompt effectiveness, and output formatting. Validation includes toxicity scanning and enforcement of naming, formatting, and size constraints to ensure safety and consistency.

Agent configurations specify trigger conditions in the frontmatter to determine when the agent activates. Design these conditions to match your plugin's use case, ensuring the agent responds only when appropriate contexts are detected.

Include 2–4 example blocks that demonstrate expected agent behavior for common inputs and scenarios. Examples should show how the agent interprets requests, applies its system prompt, and formats outputs, providing templates for consistent agent performance.

Agent configuration allows you to specify tools within the frontmatter to extend agent capabilities. Define tools aligned with your agent's responsibilities; validation ensures they integrate correctly with the agent's system prompt and trigger conditions.