Self-Learn Skill

Two-part continuous improvement loop: learn from corrections, learn from self-evaluation.

On Every Task Completion

After finishing any task, briefly self-evaluate (1–3 sentences internally):

• Did the output meet the intent?
• What could be better next time?
• Is this a repeatable pattern worth storing?

If yes → log it. If it was routine with no issues → skip logging (don't over-log).

Log format:

CONTEXT: [task type]
REFLECTION: [what I noticed]
LESSON: [what to do differently / keep doing]

Append to memory/corrections.md under today's date. Also call memory_store with:

• category: decision
• importance: 0.75
• text: [LESSON] <lesson text> + relevant keywords

On User Corrections

Trigger phrases (detect these):

• "No, that's wrong / not right"
• "Actually..." / "I prefer..." / "Remember that I..."
• "Stop doing X" / "Why do you keep..."
• "I told you before..." / "Always do X"

When triggered:

1. Acknowledge the correction briefly
2. Append to memory/corrections.md under ## Corrections with today's date
3. Call memory_store with:
   • category: preference (style/tone) or decision (behaviour/approach)
   • importance: 0.85
   • text: [CORRECTION] <what was wrong> → <correct behaviour> + keywords
4. Recall to verify it stored correctly

Log format:

[YYYY-MM-DD] CORRECTION: <what was wrong> → <correct behaviour>

memory/corrections.md Structure

# Corrections & Learnings Log

## Corrections
[YYYY-MM-DD] CORRECTION: ...

## Self-Evaluations
[YYYY-MM-DD] CONTEXT: ... | LESSON: ...

Create the file if it doesn't exist.

On "What have you learned?" / "Show my patterns"

Read memory/corrections.md and show last 10 entries. Also memory_recall with query "CORRECTION OR LESSON" for LanceDB results.

Installation on a New OpenClaw

1. Copy skills/self-learn/ into your workspace skills/ folder — skill activates automatically
2. Create memory/corrections.md (copy from references/corrections-template.md)
3. Optionally update AGENTS.md skill table for easy reference

Rules

• Don't over-log — skip routine tasks with no notable outcome
• Atomic entries — one lesson per entry, under 100 words
• Keywords matter — include domain keywords in memory_store text for good recall
• No secrets — never log credentials, personal data, or sensitive info
• Corrections always log — user corrections are always worth storing (importance ≥ 0.85)