Feature Development

Lightweight workflow for simple features with built-in reflection, testing, code review, and commit stages.

For complex features requiring architecture planning, use Spec-Driven Development workflow.

When to Use

  • Simple features and enhancements

  • Bug fixes and small improvements

  • Refactoring existing code

  • Changes that do not require architectural decisions

Plugins needed for this workflow

Workflow

How It Works

1. Implement the feature

Start by implementing your feature. Describe what you want to build and let the LLM write the code.

After LLM completes, review the generated code to ensure it addresses your requirements before proceeding.

2. Reflect on implementation

Use the /reflexion:reflect command to have the LLM review its own work, identify potential issues, and suggest improvements.

After LLM completes, it will apply improvements automatically. If you want deeper analysis from multiple perspectives, use /reflexion:critique instead.

3. Write tests

Use the /tdd:write-tests command to generate tests covering the changes you made. You can optionally specify areas to focus on.

Or with specific focus areas:

After LLM completes, verify that all tests pass. If tests fail, ask the LLM to fix the issues before continuing.

4. Review local changes

Use the /code-review:review-local-changes command to run a comprehensive multi-agent code review on your uncommitted changes.

After LLM completes, review the findings organized by severity (Critical, High, Medium, Low). Address Critical and High priority issues before committing. You can ask the LLM to fix specific issues.

5. Preserve learnings

Use the /reflexion:memorize command to save valuable insights and patterns discovered during development to your project memory.

Or with specific context:

After LLM completes, the insights are saved to CLAUDE.md, making them available for future development sessions.

6. Create commit

Use the /git:commit command to create a well-formatted conventional commit with appropriate emoji.

After LLM completes, a commit is created with a descriptive message following conventional commit format. You can then push your changes or create a pull request using /git:create-pr.

Tips

  • Skip steps when appropriate: For trivial changes, you may skip reflection or memorization

  • Iterate when needed: Run reflect and review multiple times for complex changes

  • Fix before commit: Always address Critical and High priority review findings before committing

  • Be specific: Provide context to commands for better results (e.g., focus areas for tests)

Last updated