cursor-subagent-creator

What it does

The skill defines a procedure to create Cursor-specific subagents that operate with isolated context for multi-step workflows. It instructs when to use these subagents (for new subagents, specialized assistants, or complex workflows) and clarifies that they should follow Cursor patterns and be placed under .cursor/agents/ in projects or ~/.cursor/agents/ for users. It explains the subagent structure, including the required frontmatter fields (name, description, model, readonly, is_background) and the file format for subagents. It outlines a creation process emphasizing purpose, location, and frontmatter configuration, and provides structured prompts for the subagent content. It includes examples for common subagent types (verifier, debugger, security-auditor, test-runner, doc-writer, orchestrator) and practical usage patterns (automatic delegation, explicit invocation, parallel execution, resuming). It also prescribes best practices, formatting, and a quick template for new subagents, along with a quality checklist to ensure specificity, actionable instructions, and proper configuration.

How it works

1. Use when the user requests creation of a new subagent or when complex, multi-step workflows require isolated context.
2. Follow the Subagent Creation Process: define purpose, choose location, configure frontmatter (name, description, model, readonly, is_background), write the subagent prompt with identity, invocation context, process steps, and expected output, and ensure focus and specificity.
3. Adhere to the Frontmatter and file structure conventions, naming in kebab-case, and described placement paths.
4. Utilize provided patterns for common subagents (verifier, debugger, security-auditor, test-runner, doc-writer, orchestrator) as templates or references.
5. Implement automatic delegation cues and explicit invocation syntax for interacting with the subagent, including the /name and natural-language methods.
6. Ensure output is structured and testable, and maintain best practices and a concise output format.

When to use it

Use this skill when the user asks to:

• Create a new subagent/agent
• Create a specialized assistant
• Implement a complex workflow with multiple steps
• Create verifiers, auditors, or domain experts
• Tasks that require isolated context and multiple steps

What it can touch

• Subagents are defined with a markdown file in .cursor/agents/ (project) or ~/.cursor/agents/ (user).
• The skill references fields in frontmatter and a specific Markdown subagent prompt format.
• It mentions commands and tools like You are an [expert in X] and the structured output/report formats described in examples, but does not execute external tools directly in this summary.

Caveats

• The description emphasizes focused, single-responsibility subagents and discourages vague prompts.
• It requires kebab-case naming and explicit frontmatter fields.
• It advises selecting appropriate model settings (inherit, fast, or specific IDs) and optional readonly/is_background flags.
• License is NOASSERTION per the source metadata.