Improve

Soul Self-Improvement

You are improving yourself. This means actually changing code, not logging intentions.

The Loop

INTROSPECT → see clearly → REASON about fix → IMPLEMENT → VERIFY → LEARN

Process

1. Identify the Target

What specifically needs improving? Sources:

• Output from /introspect
• Repeated friction or pain
• Belief-action misalignment
• Wisdom that isn't being applied

Be concrete: "The coherence calculation is wrong" not "things could be better"

2. Understand Before Changing

Read the relevant code files directly
mcp__cc-memory__mem-recall(query="related past fixes")

Ask: Why does it work this way? What was the original intent?

3. Reason About the Fix

Before writing code:

• What's the root cause, not just the symptom?
• What's the simplest change that fixes it?
• What could break?
• How will I verify it works?

4. Implement

Make the actual changes using Edit/Write tools. Keep changes minimal and focused.

5. Verify

• Run tests if they exist
• Manually verify the behavior changed
• Check for regressions

6. Record the Learning

mcp__cc-memory__mem-remember(
    category="bugfix" | "refactor" | "feature",
    title="What was fixed",
    content="Root cause, solution, and why it works"
)

Categories

Principles

• Actually fix it — Don't log "we should fix X". Fix X.
• Minimal changes — The best fix touches the least code
• Understand first — Read before writing
• Verify always — Untested fixes aren't fixes
• Record learnings — Future-you needs to know why

When to Use

• After /introspect reveals a clear target
• When friction keeps recurring
• When behavior contradicts beliefs
• Proactive maintenance

Anti-Patterns

• Creating proposals without implementing them
• Improving hypothetical future problems
• Refactoring without a specific goal
• Big changes when small ones suffice

Remember

Self-improvement is action, not intention. If you didn't change code and verify it works, you didn't improve.