Overview

Provide a lightweight, heartbeat-driven self-reflection and mistake-tracking system for OpenClaw agents using simple local files and CLI commands.

Key Advantages

1.Enables continuous improvement by regularly prompting for reflection based on a time threshold.

2.Simple, Unix-style CLI with clear subcommands: check, log, read, stats, reset.

3.Persists lessons learned in a configurable memory file for later recall and context.

4.Integrates cleanly with OpenClaw heartbeat configuration using standard JSON config files.

5.Provides basic analytics (stats) to visualize reflection frequency and coverage over time.

Use Cases

Individual developers tracking recurring mistakes (e.g., API timeouts, error-handling gaps) and their corresponding fixes.
Long-running OpenClaw agents that should periodically review recent errors and update their operating heuristics.
Engineering teams that want a lightweight process layer for postmortem-style notes without a full incident management system.
Personal productivity or learning workflows where the agent logs insights, misses, and corrections throughout the day.
Experimental setups where one wants to observe whether reflection prompts reduce repeated errors over time.

Evaluation Scores

7.9

/ 10

Reliability

7.5

Functionality

8.0

Usability

7.8

Safety

7.5

Performance

9.0

Compatibility

8.5

Based on 1 evaluation · Latest: 3/19/2026

Download Trend

Loading...

Evaluation History (1)

7.9/103/19/2026darwin-arm64

▼

OS: darwin-arm64LLM: minimax/minimax-m2.5

**Judgment:** A focused, well-scoped skill that adds structured self-reflection and error-tracking to OpenClaw agents with minimal overhead. Best suited to users comfortable with editing config files and working with CLI tools. **What it does well** - Implements a clear heartbeat-driven loop: periodically check if reflection is due, then read prior lessons and log new ones. - Uses simple local files (`memory_file`, `state_file`) for persistence, making it easy to inspect, back up, or version-control reflections. - Offers practical commands (`check`, `log`, `read`, `stats`, `reset`) that map directly to reflection workflows. - Integrates with standard OpenClaw configuration (`~/.openclaw/openclaw.json` and `HEARTBEAT.md`). **Key risks / limitations** - **Data sensitivity:** Reflections are stored in plain text; logging sensitive data (credentials, PII, proprietary details) is risky if the workspace is shared or synced to remote repos. - **Human-in-the-loop dependency:** The system prompts reflection but does not enforce quality; if users log vague or low-effort entries, the value of the memory file degrades. - **Manual configuration:** Requires editing JSON configs and maintaining `HEARTBEAT.md`; non-technical users may struggle to set it up correctly. - **No built-in semantic analysis:** It records and replays lessons but does not automatically generalize patterns or detect duplicates. **Recommended scenarios** - Long-lived development agents where preventing repeat mistakes (like missing timeouts or mishandling edge cases) matters. - Developers or teams who already use Git and config files and want a minimal reflection layer without extra infrastructure. - Personal learning or productivity setups where the user periodically pauses to capture “miss + fix + tag” style reflections. **Less ideal for** - Highly regulated or sensitive environments unless combined with strict data-handling policies. - Non-technical users who are uncomfortable editing JSON config files or working via the command line. - Use cases requiring automated analysis or optimization of reflection content (this skill focuses on structure and storage, not AI-driven analysis).

Self Reflection