create-subagent — скилл для Claude Code: что делает и как установить
Скилл create-subagent помогает создавать изолированных субагентов с готовыми системными промптами для ревью кода, отладки и работы с данными.
MIT ★ 328 обновлён 2026-10-29
npx skills add coco-research/coco --skill create-subagent -a claude-codeЧто делает create-subagent
Скилл помогает спроектировать и записать конфигурации субагентов в формате Markdown. Он задаёт разметку .md-файла и формулирует системный промпт под задачи ревью, тестирования или анализа данных.
Файлы размещаются на двух уровнях: в каталоге .cursor/agents/ для текущего репозитория или в ~/.cursor/agents/ для всех проектов сразу. При совпадении имён проектная конфигурация имеет приоритет перед пользовательской. В заголовке обязательны поля name строчными латинскими буквами через дефис и description, по которому модель определяет момент делегирования задачи.
Если в описании пропустить чёткие триггеры вызова или фразу «use proactively», агент перестанет запускаться автоматически.
Какие задачи решает
- изолировать глубокий поиск по коду от контекста основного диалога
- настроить проектного ревьюера с чётким чек-листом для коммитов
- создать системный промпт для отладчика ошибок и стектрейсов
- вынести общих персональных ассистентов в глобальную папку пользователя
Примеры запросов агенту
- Помоги создать субагента для ревью кода в папке
.cursor/agents/ - Сделай субагента отладчика, который ищет первопричину багов и правит код
- Настрой глобального субагента data-scientist для написания SQL-запросов
Как установить create-subagent
1Через Skills CLI (Vercel)
npx skills add coco-research/coco --skill create-subagent -a claude-code2Вручную: положить файлы
~/.claude/skills/create-subagent/SKILL.mdПапка скилла целиком; для одного проекта — .claude/skills/create-subagent/
1Через Skills CLI (Vercel)
npx skills add coco-research/coco --skill create-subagent -a codex2Вручную: положить файлы
.agents/skills/create-subagent/SKILL.mdДля всех проектов — ~/.agents/skills/
1Через Skills CLI (Vercel)
npx skills add coco-research/coco --skill create-subagent -a cursor2Вручную: положить файлы
.cursor/skills/create-subagent/SKILL.mdCursor читает и .agents/skills/, и .claude/skills/
1GitHub CLI
gh skill install coco-research/coco adapters/cursor/skills/create-subagent2Вручную: положить файлы
.github/skills/create-subagent/SKILL.md1Через Skills CLI (Vercel)
npx skills add coco-research/coco --skill create-subagent -a gemini-cli2Командой платформы
gemini skills install https://github.com/coco-research/coco --path adapters/cursor/skills/create-subagentВручную: положить файлы
.devin/skills/create-subagent/SKILL.mdСтарый путь .windsurf/skills/ тоже читается
Claude.ai
Упакуйте папку в zip и загрузите в Settings → Capabilities → Skills
По шагам, с командами для каждой платформы: Как установить скилл в Claude Code: пять способов
Вопросы о create-subagent
Как установить create-subagent в Claude Code?
Выполните команду npx skills add coco-research/coco --skill create-subagent -a claude-code.
Зачем нужен скилл create-subagent?
Он помогает составить системный промпт и frontmatter для специализированного помощника, чтобы изолировать контекст задачи от основного чата.
Где сохраняются созданные субагенты?
Проектные агенты сохраняются в папке .cursor/agents/, а персональные конфигурации — в каталоге ~/.cursor/agents/.
Исходный текст
SKILL.md · MIT
--- name: create-subagent description: Create custom subagents for specialized AI tasks. Use when the user wants to create a new type of subagent, set up task-specific agents, configure code reviewers, debuggers, or domain-specific assistants with custom prompts. disable-model-invocation: true --- # Creating Custom Subagents This skill guides you through creating custom subagents for Cursor. Subagents are specialized AI assistants that run in isolated contexts with custom system prompts. ## When to Use Subagents Subagents help you: - **Preserve context** by isolating exploration from your main conversation - **Specialize behavior** with focused system prompts for specific domains - **Reuse configurations** across projects with user-level subagents ### Inferring from Context If you have previous conversation context, infer the subagent's purpose and behavior from what was discussed. Create the subagent based on specialized tasks or workflows that emerged in the conversation. ## Subagent Locations | Location | Scope | Priority | |----------|-------|----------| | `.cursor/agents/` | Current project | Higher | | `~/.cursor/agents/` | All your projects | Lower | When multiple subagents share the same name, the higher-priority location wins. **Project subagents** (`.cursor/agents/`): Ideal for codebase-specific agents. Check into version control to share with your team. **User subagents** (`~/.cursor/agents/`): Personal agents available across all your projects. ## Subagent File Format Create a `.md` file with YAML frontmatter and a markdown body (the system prompt): ```markdown --- name: code-reviewer description: Reviews code for quality and best practices --- You are a code reviewer. When invoked, analyze the code and provide specific, actionable feedback on quality, security, and best practices. ``` ### Required Fields | Field | Description | |-------|-------------| | `name` | Unique identifier (lowercase letters and hyphens only) | | `description` | When to delegate to this subagent (be specific!) | ## Writing Effective Descriptions The description is **critical** - the AI uses it to decide when to delegate. ```yaml # BAD: Too vague description: Helps with code # GOOD: Specific and actionable description: Expert code review specialist. Proactively reviews code for quality, security, and maintainability. Use immediately after writing or modifying code. ``` Include "use proactively" to encourage automatic delegation. ## Example Subagents ### Code Reviewer ```markdown --- name: code-reviewer description: Expert code review specialist. Proactively reviews code for quality, security, and maintainability. Use immediately after writing or modifying code. --- You are a senior code reviewer ensuring high standards of code quality and security. When invoked: 1. Run git diff to see recent changes 2. Focus on modified files 3. Begin review immediately Review checklist: - Code is clear and readable - Functions and variables are well-named - No duplicated code - Proper error handling - No exposed secrets or API keys - Input validation implemented - Good test coverage - Performance considerations addressed Provide feedback organized by priority: - Critical issues (must fix) - Warnings (should fix) - Suggestions (consider improving) Include specific examples of how to fix issues. ``` ### Debugger ```markdown --- name: debugger description: Debugging specialist for errors, test failures, and unexpected behavior. Use proactively when encountering any issues. --- You are an expert debugger specializing in root cause analysis. When invoked: 1. Capture error message and stack trace 2. Identify reproduction steps 3. Isolate the failure location 4. Implement minimal fix 5. Verify solution works Debugging process: - Analyze error messages and logs - Check recent code changes - Form and test hypotheses - Add strategic debug logging - Inspect variable states For each issue, provide: - Root cause explanation - Evidence supporting the diagnosis - Specific code fix - Testing approach - Prevention recommendations Focus on fixing the underlying issue, not the symptoms. ``` ### Data Scientist ```markdown --- name: data-scientist description: Data analysis expert for SQL queries, BigQuery operations, and data insights. Use proactively for data analysis tasks and queries. --- You are a data scientist specializing in SQL and BigQuery analysis. When invoked: 1. Understand the data analysis requirement 2. Write efficient SQL queries 3. Use BigQuery command line tools (bq) when appropriate 4. Analyze and summarize results 5. Present findings clearly Key practices: - Write optimized SQL queries with proper filters - Use appropriate aggregations and joins - Include comments explaining complex logic - Format results for readability - Provide data-driven recommendations For each analysis: - Explain the query approach - Document any assumptions - Highlight key findings - Suggest next steps based on data Always ensure queries are efficient and cost-effective. ``` ## Subagent Creation Workflow ### Step 1: Decide the Scope - **Project-level** (`.cursor/agents/`): For codebase-specific agents shared with team - **User-level** (`~/.cursor/agents/`): For personal agents across all projects ### Step 2: Create the File ```bash # For project-level mkdir -p .cursor/agents touch .cursor/agents/my-agent.md # For user-level mkdir -p ~/.cursor/agents touch ~/.cursor/agents/my-agent.md ``` ### Step 3: Define Configuration Write the frontmatter with the required fields (`name` and `description`). ### Step 4: Write the System Prompt The body becomes the system prompt. Be specific about: - What the agent should do when invoked - The workflow or process to follow - Output format and structure - Any constraints or guidelines ### Step 5: Test the Agent Ask the AI to use your new agent: ``` Use the my-agent subagent to [task description] ``` ## Best Practices 1. **Design focused subagents**: Each should excel at one specific task 2. **Write detailed descriptions**: Include trigger terms so the AI knows when to delegate 3. **Check into version control**: Share project subagents with your team 4. **Use proactive language**: Include "use proactively" in descriptions ## Troubleshooting ### Subagent Not Found - Ensure file is in `.cursor/agents/` or `~/.cursor/agents/` - Check file has `.md` extension - Verify YAML frontmatter syntax is valid
© coco-research · MIT · coco-research/coco