reflect

What This Skill Does

The reflect skill is the foundational memory engine for the OpenClaw ecosystem, designed to transform fleeting chat interactions into persistent, actionable intelligence. It operates by managing skill-specific MEMORY.md files that act as a long-term knowledge base for the AI agent. Instead of relying on a sliding window of context or repeating instructions across different sessions, the reflect skill enables the agent to store user preferences, coding patterns, successful debugging outcomes, and corrected behaviors in a structured, hierarchical manner. When activated, it facilitates a feedback loop where the agent synthesizes user feedback into durable rules, ensuring that every project-specific constraint or stylistic preference is remembered and applied in subsequent interactions.

Installation

To integrate the reflect skill into your environment, use the following OpenClaw command in your terminal:

clawhub install openclaw/skills/skills/anton-abyzov/sw-reflect

Once installed, ensure your environment variables are configured correctly to detect your project root so the agent can properly map skill-specific memory files to the appropriate directory structure.

Use Cases

• Coding Standards Enforcement: Automatically remember that a specific project prefers functional programming patterns over object-oriented ones or uses specific library versions.
• Preference Persistence: Train the agent on your preferred tone, response length, or documentation style so you don't have to re-instruct it on "how to talk" every time you start a new chat.
• Error Resolution: Capture the specific steps or workarounds used to resolve obscure system bugs so that the agent can proactively suggest the fix if the issue recurs.
• Continuous Refinement: Use it as a post-task feedback tool to log what worked and what didn't in complex workflows like automated deployment or data parsing.

Example Prompts

1. "Reflect on our work today: we found that using React Query is more efficient for our API calls than raw useEffect hooks. Add this as a rule to the frontend memory."
2. "Whenever I ask you to write unit tests, please follow the specific pattern we established in the last session for our integration tests. Save this to the testing memory."
3. "Reflect: From now on, always ensure all generated documentation files include a brief summary section at the top. Store this preference."

Tips & Limitations

• Synthesize, Don't Copy: The reflect skill works best when the agent distills raw dialogue into concise, actionable instructions rather than dumping long chat transcripts into the memory files.
• Hierarchy Matters: Understand that the system uses a tiered approach (Skill Memory -> Project Memory -> Global Memory). Always ensure your feedback is targeted to the correct level to avoid cluttering global settings with project-specific logic.
• Maintainability: Periodically review your MEMORY.md files to ensure they remain relevant; as projects evolve, some "learned" rules may eventually become obsolete and should be pruned to prevent conflicting instructions.