design-agent

Designs plugin agent specifications with frontmatter, skill assignments, execution flows, and operating principles.

1|Updated Mar 22, 2026
One-click install
npx skills add https://github.com/JaviMontano/mao-plugin-qa --skill design-agent-javimontano
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: design-agent
Source: https://github.com/JaviMontano/mao-plugin-qa/tree/main/skills/design-agent
Command: npx skills add https://github.com/JaviMontano/mao-plugin-qa --skill design-agent-javimontano

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Designing agents for Claude Code plugins requires knowing which frontmatter fields are allowed, how to assign skills, and how to structure execution flows. This Skill produces a complete, constraint-compliant agent design document so you avoid runtime errors from forbidden fields like hooks, mcpServers, or permissionMode. ## Core Features & Use Cases - Constraint Enforcement: Validates that plugin subagent designs exclude forbidden frontmatter fields (hooks, mcpServers, permissionMode) using a reference constraint matrix. - Structured Design Workflow: Guides role definition, tool allowlist selection, skill mapping, execution flow documentation, and maxTurns calculation across eight steps. - Operating Principles Authoring: Helps write specific, actionable, and verifiable behavioral rules for the agent rather than generic guidance. - Use Case: When building a multi-agent QA plugin, use this Skill to draft the orchestrator agent's full specification—including skill tables, Mermaid execution flows, and delegation criteria—before generating the deployable agent file. ## Quick Start Ask the assistant to design an agent named plugin-qa-engineer for your plugin, including its role, tools, assigned skills, and execution flows.

Frequently Asked Questions about design-agent

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I design an agent for a Claude Code plugin?▼

Define the agent's role and boundary, draft YAML frontmatter with name, description, and tools, map assigned skills, document execution flows per command, and write operating principles. The design is produced as a specification document for review before creating the deployable agent file.

What frontmatter fields are forbidden in plugin subagents?▼

Plugin subagents cannot use hooks, mcpServers, or permissionMode in their frontmatter. Hooks are silently ignored, mcpServers cause runtime errors, and permissionMode is overridden by session-level permissions. Use the plugin's hooks.json for lifecycle automation instead.

Should I use tools or disallowedTools in agent frontmatter?▼

Use either tools (allowlist) or disallowedTools (blocklist), never both. If both are present, disallowedTools is silently ignored. The allowlist approach follows least-privilege principles and is recommended for plugin agents.

How is maxTurns calculated for a plugin agent?▼

Estimate maxTurns with the formula: (number of skills × 4) + complexity bonus + (interaction points × 2), rounded up to the nearest 5. This is a heuristic; actual turn usage depends on task complexity and user interaction at runtime.

What are the limitations of plugin subagent design?▼

The skill produces a design document, not the final deployable file. Operating principles are prompt-level guidance, not enforced constraints, and agents cannot invoke skills from other plugins. Skill assignments assume the architecture plan is finalized.