agent-development
This skill should be used when the user asks to "create an agent", "add an agent", "write a subagent", "agent frontmatter", "when to use description", "agent examples", "agent tools", "agent colors", "autonomous agent", or needs guidance on agent structure, system prompts, triggering conditions, or agent development best practices for Claude Code plugins.
Security Assessment
About agent-development
The agent-development skill is designed to guide users in creating, managing, and optimizing autonomous agents within Claude Code plugins. It addresses the complexity of structuring agents, defining system prompts, setting triggering conditions, and selecting appropriate models and tools. By providing a clear framework for agent creation, this skill ensures that agents can operate independently on multi-step tasks while aligning with user intent and plugin requirements.
Key features include detailed guidance on the agent file structure, frontmatter requirements, and best practices for specifying triggering conditions and responsibilities. Users can customize agents with specific models, colors, and tool access, ensuring both functional and visual clarity within a plugin. The skill emphasizes the distinction between autonomous agent actions and user-initiated commands, promoting efficient task delegation and robust agent behavior.
This skill is particularly useful for developers and technical writers creating Claude Code plugins who need structured guidance on agent development. Typical use cases include defining agents for code review, documentation generation, testing, or analysis tasks. By following the prescribed file formats and frontmatter conventions, users can ensure consistent agent behavior, simplify maintenance, and improve the overall effectiveness of autonomous operations within their applications.
FAQ
How do I define when an agent should be triggered?
You should specify triggering conditions in the 'description' frontmatter field, including 2-4 typical scenarios in prose and a pointer to detailed scenarios under the 'When to invoke' section in the agent body.
Which models can agents use?
Agents can use 'inherit' to follow the parent plugin's model, or specific models like 'sonnet', 'opus', and 'haiku' depending on capability needs. 'Inherit' is recommended for most cases.
Can I limit which tools an agent can access?
Yes, the optional 'tools' frontmatter field allows you to restrict an agent to specific tools such as 'Read', 'Write', or 'Grep'.
What are the guidelines for choosing an agent color?
Colors help visually distinguish agents: blue/cyan for analysis, green for success-oriented tasks, yellow for caution/validation, red for critical/security, and magenta for creative tasks. Consistency across similar agents is recommended.
Are there restrictions on agent names?
Yes, agent names must be lowercase, 3-50 characters, only use letters, numbers, and hyphens, and start and end with an alphanumeric character. Underscores or generic names should be avoided.
Install agent-development
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
anthropics/claude-plugins-official