auto-memory-pro

What This Skill Does

Auto-Memory-Pro is a sophisticated intelligence layer designed to evolve alongside your project. While Claude Code’s built-in memory automatically tracks project patterns and session insights in MEMORY.md, these notes often become cluttered, repetitive, or difficult to manage over time. Auto-Memory-Pro automates the curation process, transforming raw logs into high-value project documentation. It identifies recurring patterns, validates successful debugging solutions, and promotes these insights into enforced CLAUDE.md rules or standalone reusable skills. By acting as a gardener for your project's institutional knowledge, it ensures that your AI agent remains consistent, efficient, and increasingly capable of handling project-specific nuances without re-learning the same lessons.

Installation

Installation is straightforward depending on your preferred environment. For OpenClaw users, simply execute the following command in your terminal: clawhub install openclaw/skills/skills/alirezarezvani/auto-memory-pro

If you are integrating via the Codex CLI, use: ./scripts/codex-install.sh --skill self-improving-agent. For standard Claude Code environments, ensure you add the plugin via: /plugin marketplace add alirezarezvani/claude-skills followed by /plugin install self-improving-agent@claude-code-skills.

Use Cases

This skill is ideal for projects where project velocity is high and architectural standards are constantly evolving. Key use cases include:

1. Refining Project Standards: Automatically moving tribal knowledge from scratchpads into your formal CLAUDE.md file.
2. Debugging Efficiency: Converting a complex one-time fix for a recurring bug into an automated rule, ensuring the agent doesn't repeat past mistakes.
3. Knowledge Cleanup: Periodic pruning of MEMORY.md to ensure the most relevant context is kept within the first 200 lines for optimal agent performance.
4. Onboarding Context: Preparing a project for a new team member by condensing fragmented memory into clear, enforced rule sets.

Example Prompts

1. "Review my MEMORY.md and suggest which recurring patterns should be graduated to our .claude/rules/ directory for better enforcement."
2. "I just fixed a tricky dependency conflict; use the /si:extract command to create a reusable skill so Claude handles this configuration automatically in the future."
3. "Run a health check on my project memory and show me which topics are stale or taking up unnecessary context space."

Tips & Limitations

• Be Selective: Use /si:promote for stable architectural patterns, but avoid promoting experimental code that might change frequently.
• Hierarchy Matters: Always prioritize formalizing rules in .claude/rules/ for file-specific behavior, as these provide the most surgical guidance for the agent.
• Review Periodically: Even though the process is automated, always perform a final visual inspection of promoted rules to ensure the generated documentation aligns with your current team standards.
• Limit: The skill relies on the agent's ability to interpret MEMORY.md. If your memory file has grown massive and lacks structure, the first analysis pass might require your manual guidance to help the agent focus on key areas.