/build
Implement a task directly without creating a plan first.
How it fires
How this command gets triggered: by you, by Claude, or both.
- Fires itselfClaude auto-loads it when your prompt matches the work.
- You can call itInvoke it directly when you want it.
- Slash command
/build
Context preview
What this command does when you run it.
Implement a task directly without creating a plan first.
Command definition
build.mdBuild Task
Implement a task directly without creating a plan first.
Variables
adw_id: $ARGUMENTS task_description: $ARGUMENTS
Instructions
1. **Prime with Context**: First, read and execute `.claude/commands/prime.md` to understand the codebase 2. **Analyze Task**: Carefully read and understand the task description 3. **Implement Solution**: Think hard and then directly implement the solution for the task 4. **Validate Work**: Ensure the implementation is complete and working 5. **Report Results**: Summarize what was done
Setup Phase
Before implementing the task:
- Execute the prime command to understand the codebase structure
- Read relevant documentation files (README.md, etc.)
- Understand the existing patterns and conventions
Implementation Guidelines
- Follow existing code patterns and conventions
- Use the libraries and frameworks already in the codebase
- Write clean, maintainable code
- Add appropriate error handling
- Follow security best practices
Task Description
task_description
Expected Actions
1. **Research**: Understand the codebase and task requirements 2. **Implement**: Make the necessary changes to complete the task 3. **Test**: Verify the implementation works as expected 4. **Commit**: Create a meaningful commit with the changes
Report
After completing the implementation:
- Summarize the work done in clear bullet points
- List all files created or modified
- Report the total lines changed with `git diff --stat`
- Note any important decisions or trade-offs made
- Highlight any follow-up tasks that may be needed
Read more
Build Task
Implement a task directly without creating a plan first.
Variables
adw_id: $ARGUMENTS task_description: $ARGUMENTS
Instructions
1. **Prime with Context**: First, read and execute `.claude/commands/prime.md` to understand the codebase 2. **Analyze Task**: Carefully read and understand the task description 3. **Implement Solution**: Think hard and then directly implement the solution for the task 4. **Validate Work**: Ensure the implementation is complete and working 5. **Report Results**: Summarize what was done
Setup Phase
Before implementing the task:
- Execute the prime command to understand the codebase structure
- Read relevant documentation files (README.md, etc.)
- Understand the existing patterns and conventions
Implementation Guidelines
- Follow existing code patterns and conventions
- Use the libraries and frameworks already in the codebase
- Write clean, maintainable code
- Add appropriate error handling
- Follow security best practices
Task Description
task_description
Expected Actions
1. **Research**: Understand the codebase and task requirements 2. **Implement**: Make the necessary changes to complete the task 3. **Test**: Verify the implementation works as expected 4. **Commit**: Create a meaningful commit with the changes
Report
After completing the implementation:
- Summarize the work done in clear bullet points
- List all files created or modified
- Report the total lines changed with `git diff --stat`
- Note any important decisions or trade-offs made
- Highlight any follow-up tasks that may be needed
Watched how we used GPT-5 and Claude Code with nano-agents here. What? A MCP Server for experimental, small scale engineering agents with multi-provider LLM support. Why?
Other commands on nano-agent.
- /convert_paths_absolute
Converts relative paths in .claude/settings.json command scripts to absolute paths
Open command - /convert_paths_relative
Converts absolute paths in .claude/settings.json command scripts to relative paths
Open command - /create_worktree
Create a new git worktree for an agent to work in isolation.
Open command - /hop_evaluate_nano_agents
Using the nano-agent mcp server, execute the following nano agents with their respective prompts, models, and providers then rank the results based on the `Response Format`.
Open command - /lop_eval_1__dummy_test
- Pass the prompt into each nano-agent AS IS. Do not change the prompt in any way.
Open command - /lop_eval_2__basic_read_test
- Pass the prompt into each nano-agent AS IS. Do not change the prompt in any way. - Each agent should execute the task independently.
Open command

