Continuous Improvement

After completing a task, briefly reflect and capture learnings.

When to Activate

• Task just completed
• Same mistake made twice
• User asks "what did I learn" or "retrospective"
• Complex process was difficult or slow

Reflect (30 seconds)

Ask yourself:

• What worked well?
• What was difficult or slow?
• What's missing (automation, docs, tools)?

Capture

| Learning Type | Action | |---------------|--------| | Pattern/gotcha | Store in Memory MCP | | Missing automation | Create issue/TODO | | Repeated mistake | Update .claude/rules/ | | Complex process | Create .claude/skills/ |

Memory Pattern

When storing a learning in Memory MCP:

mcp__memory__create_entities({
  entities: [{
    name: "Learning_TOPIC",
    entityType: "development_learning",
    observations: [
      "Context: What was happening",
      "Learning: What was learned",
      "Application: How to apply it"
    ]
  }]
})

Improvement Actions

| Issue | Solution | |-------|----------| | Same mistake twice | Add rule to .claude/rules/ | | Manual process > 5 min | Create skill or hook | | New tool/pattern | Store in Memory MCP | | Missing documentation | Update docs |

Rule Addition Example

When you make the same mistake twice:

1. Identify the pattern
2. Create or update a rule file:

# .claude/rules/my-rules.md
---
paths: relevant/path/**
description: Rules for this area
---

## Critical Rules

| Rule | Why |
|------|-----|
| Don't do X | Because Y happens |

Skill Addition Example

When a process is complex and repeated:

1. Create skill directory: .claude/skills/my-skill/
2. Create SKILL.md:

# .claude/skills/my-skill/SKILL.md
---
name: my-skill
description: When this skill should activate
---

# My Skill

## When to Activate
[Trigger conditions]

## Process
[Steps to follow]

Run /improve-claude-config for major configuration updates.