Self Refine Skill — AI skill for Claude Code
🪞 Systematic self-reflection for AI agents.
How to install Self Refine Skill
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open Thomaszhou22/self-refine-skill and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Self Refine Skill does
🪞 Systematic self-reflection for AI agents. GENERATE → CRITIQUE → REFINE → CHECK. Based on peer-reviewed research. Zero API cost. Works with Claude Code, Cursor, Copilot, Codex CLI, and 10+ more platforms.
Alternatives in AI
- Manifest — Real-time cost observability for OpenClaw agents — track tokens, costs, messages, and model usage 4.1k ★
- WindsurfAPI — Turn Windsurf / Devin Desktop's 100+ AI models (Claude, GPT, Gemini, DeepSeek, Kimi, GLM, SWE) into OpenAI-, A 3k ★
- Ante — Ghost in your shell 1.9k ★
README
🪞 Self-Refine Skill
**Systematic self-reflection for AI agents.**
*Make every AI response better — before the user sees it.*
[](LICENSE) [](https://openclaw.ai) [](https://claude.ai) [](https://cursor.sh) [](https://github.com/features/copilot) [](https://github.com/openai/codex) [](https://ai.google.dev) [](https://codeium.com/windsurf) [](https://www.jetbrains.com/ai/) [](BENCHMARK.md)
What It Does
Most AI responses are first drafts. This skill adds an **invisible self-review step** before every response:
GENERATE → CRITIQUE → REFINE → CHECK → DELIVERBased on peer-reviewed research showing **~20% quality improvement** from iterative self-refinement (Madaan et al., 2023).
**One file. Zero dependencies. Zero API cost.** The host AI acts as its own critic.
Proven Results
Benchmarked on **20 questions across 5 categories** with controlled before/after evaluation. → [Full report](BENCHMARK.md)
| Metric | Result |
|---|---|
| Overall improvement | +21.2% |
| Questions improved | 20/20 (100%) |
| Best single improvement | +43.9% |
**Per-dimension improvement:**
| Dimension | Δ |
|---|---|
| Conciseness | +27.7% |
| Actionability | +25.0% |
| Response Completeness | +24.8% |
| L |
Related Skills
Humanizer Pro
Enables LLMs to write text that passes AI detection by applying 39 composition constraints from Wikipedia and
Post Task Retro
USE WHEN any task or sweep is complete. Writes a structured self-reflection (Reflexion Q1–Q6 + AI-automation a
Research Agent Skills
AI agent skills by BESSER-PEARL, systematic paper review, and more. Works with Claude Code, Cursor, Copilot, W
Bullpen Coach
Switch Sage (the wellness coach) on or off. When on, Sage sends proactive check-ins (rest reminders, ship cele
Apronagents
Multi-agent AI coding with a human merge gate — sandboxed git for agents, small reviewed diffs, your call on e
AutoSearch
new Self-evolving deep research plugin for Claude Code -- 32 dedicated search channels (arXiv, GitHub, Reddit,
Related Agents
Literature Expert
Expert literature discovery and citation metadata agent for peer-reviewed papers, preprints, DOI lookup, and e
Async Code
Use Claude Code or CodeX CLI to perform multiple tasks in parallel with a Codex-style UI, functioning as a per
Monk Installer
Install, upgrade, and troubleshoot monk-agent plus Monk CLI/daemon for Claude, Codex, and Cursor users.
