Create Claude Code sub-agents with frontmatter, tool configuration, and system prompts.
npx skills add https://github.com/cskiro/annex --skill sub-agent-creator
Please help me install this Agent Skill. Skill: sub-agent-creator Source: https://github.com/cskiro/annex/tree/main/skills/sub-agent-creator Command: npx skills add https://github.com/cskiro/annex --skill sub-agent-creator
This Skill guides users through producing production-ready Claude Code sub-agents that conform to Anthropic's patterns, including YAML frontmatter, tool configuration, and system prompts. It streamlines creating domain-specific agents (e.g., code-reviewer, debugger) and personal or team-level agents with secure defaults.
Use this skill to generate a new sub-agent by answering a few prompts; you will get a ready-to-use Markdown file at .claude/agents/<agent-name>.md (project-level) or ~/.claude/agents/ (user-level).
High-intent search queries and answers about installing and using this skill.
Creating a Claude Code sub-agent requires defining YAML frontmatter with required fields (name, description), configuring tools with security-conscious defaults, selecting a model (inherit, sonnet, opus, haiku), and structuring a system prompt that defines role and constraints. This Skill automates that entire process, generating a production-ready Markdown file at `.claude/agents/<agent-name>.md` or `~/.claude/agents/` based on your answers to setup prompts.
Domain-specific sub-agents are Claude Code agents tailored for particular tasks—code-reviewer, debugger, data-scientist—or for proactive auto-triggering and security-sensitive workflows. Create them when you need specialized behavior, consistent tool permissions, or dedicated system prompts for repetitive tasks across projects or teams. This Skill generates them at both project and user storage levels.
Yes. This Skill supports both project-level storage (`.claude/agents/`) for project-specific agents and user-level storage (`~/.claude/agents/`) for reusable agents across projects. Automation includes frontmatter validation, tool permission setup, and system-prompt structure, so agents conform to Anthropic patterns regardless of storage location.
After generation, validate frontmatter syntax, verify tool permissions align with your security requirements, test the system prompt with sample inputs, and confirm the agent behaves as intended in your workflow. This Skill provides validation guidance and best-practices documentation to ensure production-ready quality before deployment.
No. This Skill scaffolds system-prompt structure automatically, defining role, approach, and constraints based on your agent's domain and purpose. You provide the context, and the Skill generates a template you can refine, eliminating boilerplate and ensuring consistency across agents.
This Skill supports four model-selection options: inherit (uses the parent Claude Code session's model), sonnet, opus, and haiku. Choose based on your agent's task complexity and latency requirements; the Skill applies your selection to the frontmatter and ensures correct configuration.
vinnie357Configure Claude Code agents with restricted tools and defined behaviors.
JaimeJunrInitialize an autonomous iterative task loop within the current session.
bsamieeCreate Claude Code agent files with YAML frontmatter and tool permissions.
SterlingChinScaffold new MARVIN skills with SKILL.md frontmatter and directory structures.
