reflect

What This Skill Does

The reflect skill is a meta-cognitive enhancement for your OpenClaw AI agent. It allows the agent to treat every conversation not just as a series of transient prompts, but as a source of truth for its own growth. By analyzing dialogue, the agent identifies explicit corrections, successful problem-solving paths, and architectural refinements. These insights are then permanently encoded into your agent definition files, ensuring that the same mistakes are never repeated and the same optimizations are always available. It follows the core philosophy of "Correct once, never again," transforming your agent from a static tool into an evolving partner that matures alongside your specific workflow requirements.

Installation

To add the reflection capability to your current workspace, run the following command in your terminal: clawhub install openclaw/skills/skills/stevengonsalvez/self-reflect

Once installed, you can enable the auto-reflection feature by invoking reflect on. You can verify the installation and current status of your memory optimization at any time by running reflect status.

Use Cases

• Long-term Project Context: When working on complex codebases, the agent learns your specific project style guides and library preferences, eliminating the need to repeat instructions in every session.
• Error Resolution: When you find a workaround for a misleading error message, the agent logs this as a reusable pattern, saving you minutes of investigation time in the future.
• Process Refinement: If you find that the agent is too verbose or misses specific constraints in your CLAUDE.md, the reflection skill captures these critiques and patches the configuration files automatically.

Example Prompts

1. "reflect" (Trigger an immediate analysis of the conversation history to extract persistent learnings.)
2. "reflect review" (Ask the agent to display the pending learnings identified during the last session so you can approve or reject the proposed changes.)
3. "reflect on" (Enable continuous monitoring so the agent automatically identifies and acts on improvement opportunities in real-time.)

Tips & Limitations

• High-Confidence Signals: The skill works best when you are explicit. Phrases like "always use X" or "never do Y" provide the highest confidence signals for the agent.
• Review Periodically: While the auto-reflection is powerful, use reflect review to audit the changes periodically to ensure the agent’s logic remains aligned with your intentions.
• Scope: The skill performs best when targeting structured configuration files. It is not intended to rewrite raw source code without human oversight.