reflexion
NewCollection of commands that force LLM to reflect on previous response and output. Based on papers like Self-Refine and Reflexion. These techniques improve the output of large language models by introducing feedback and refinement loops.
Summary
The reflexion skill enhances Claude Code's responses by implementing self-reflection and refinement loops, where the model critiques its own output and iteratively improves it based on feedback.
- This leads to higher quality, more accurate, and more robust code and solutions, especially for complex tasks.
Install & Usage
/plugin marketplace add <org/repo>Add the configuration to /plugin install reflexion@<marketplace>
/pluginUse Cases
Usage Examples
/reflexion Write a Python function to merge two sorted lists and then reflect on the solution for edge cases and efficiency.
/reflexion Explain the difference between a list and a tuple in Python, then refine the explanation for clarity and completeness.
/reflexion Debug this code: [paste code] and after each fix, reflect on whether the fix is correct and if there are other issues.
Security Audits
Frequently Asked Questions
What is reflexion?
The reflexion skill enhances Claude Code's responses by implementing self-reflection and refinement loops, where the model critiques its own output and iteratively improves it based on feedback. This leads to higher quality, more accurate, and more robust code and solutions, especially for complex tasks.
How to install reflexion?
To install reflexion: add a marketplace (/plugin marketplace add <org/repo>), then add the config to /plugin install reflexion@<marketplace>. Finally, /plugin in Claude Code.
What is reflexion best for?
reflexion is a plugin categorized under General. Created by Vlad Goncharov.
What can I use reflexion for?
reflexion is useful for: Refining generated code by having the model review and fix its own logic errors or inefficiencies.; Improving the clarity and completeness of documentation or explanations through iterative self-critique.; Enhancing debugging by prompting the model to reflect on potential root causes and suggest alternative fixes.; Optimizing algorithm implementations by repeatedly evaluating and refining performance and correctness.; Polishing natural language outputs like commit messages or code comments for better readability and accuracy.; Validating and improving test cases by having the model reflect on coverage and edge cases..