OpenClaw Optimizer — How-To Guide

Prerequisites: SMF Works Pro subscription active. Setup complete (see SETUP.md).

---

Table of Contents

1. How to Run a Full Workspace Audit
2. How to Analyze Context Bloat
3. How to Analyze Your Skills
4. How to Apply Optimization Recommendations
5. How to Generate and Save a Report
6. Automating with Cron
7. Combining with Other Skills
8. Troubleshooting Common Issues
9. Tips & Best Practices

---

1. How to Run a Full Workspace Audit

What this does: Checks all dimensions of your OpenClaw workspace — size, context files, skills, memory — and produces a summary with recommendations.

When to use it: Monthly, or when you notice your agent responding more slowly than usual.

Steps

Step 1 — Navigate to the skill.

cd ~/smfworks-skills/skills/openclaw-optimizer

Step 2 — Run the full audit.

python3 main.py audit

Output:

🔍 OpenClaw Workspace Audit
══════════════════════════════════════════════

Workspace: /home/user/.openclaw/workspace
Total size: 847 KB
Files: 34

📊 Context Analysis
   MEMORY.md: 24 KB ⚠️ Large — consider pruning old entries
   SOUL.md: 8 KB ✅
   AGENTS.md: 12 KB ✅
   memory/: 156 KB (18 daily files)

🔧 Skills Analysis
   Installed skills: 23 ⚠️ High count

🔴 Issues Found: 3

💡 Recommendations
   1. [HIGH] Prune MEMORY.md — 24 KB → target <10 KB
   2. [MEDIUM] Archive old daily memory files
   3. [LOW] Remove config-backup.tar.gz from workspace

Report saved to: ~/.smf/optimizer-reports/report-2024-03-15.txt

Step 3 — Prioritize the HIGH recommendations first.

Result: A clear, prioritized list of what to fix.

---

2. How to Analyze Context Bloat

What this does: Focuses specifically on files that get loaded into your agent's context window, making it slower or less focused.

When to use it: When your agent seems slower than usual or gives generic responses.

Steps

Step 1 — Run context analysis.

python3 main.py analyze --context

Output:

📊 Context Analysis

Files loaded into agent context:
  MEMORY.md: 24 KB ⚠️ Large
  SOUL.md: 8 KB ✅
  AGENTS.md: 12 KB ✅
  USER.md: 2 KB ✅
  IDENTITY.md: 1 KB ✅
  TOOLS.md: 3 KB ✅

Total context: 50 KB

Recommendation:
  MEMORY.md is the largest file (24 KB).
  Target: < 10 KB
  Action: Remove entries older than 90 days, keep only the most relevant memories

Step 2 — Prune MEMORY.md if recommended.

nano ~/.openclaw/workspace/MEMORY.md

Remove:

• Entries about completed projects you no longer need
• Outdated preferences or context
• Duplicate information
• Very old dated entries

Result: Smaller context → faster, more focused agent responses.

---

3. How to Analyze Your Skills

What this does: Shows how many skills are installed and which ones are largest.

python3 main.py analyze --skills

Output:

🔧 Skills Analysis

Installed skills: 23
Total skills size: 4.2 MB

Largest skills:
  email-campaign: 52 KB
  booking-engine: 48 KB
  report-generator: 44 KB
  form-builder: 41 KB

⚠️ High skill count (23) — Consider removing skills you don't use regularly.
   Each installed skill adds context loading overhead.

---

4. How to Apply Optimization Recommendations

What this does: Shows your current recommendations in priority order.

python3 main.py recommend

Applying each type of recommendation:

Prune MEMORY.md

nano ~/.openclaw/workspace/MEMORY.md
# Review entries, delete stale ones, save

Target: Keep only facts that are:

• Still accurate and relevant
• Not duplicated elsewhere
• Timeless rather than dated

Archive old daily memory files

mkdir -p ~/.openclaw/workspace/memory/archive
# Move files older than 30 days:
cd ~/.openclaw/workspace/memory
for f in $(ls *.md | head -n -30); do mv "$f" archive/; done

Remove unused skills

python3 ~/smfworks-skills/skills/skill-manager/main.py --list
python3 ~/smfworks-skills/skills/skill-manager/main.py --remove skill-name

Remove large non-text files from workspace

ls -lh ~/.openclaw/workspace/
# Identify files > 100KB that shouldn't be there
# Move them: mv large-file.tar.gz ~/Backups/

---

5. How to Generate and Save a Report

When to use it: To track your workspace health over time, or share the audit with someone helping you configure OpenClaw.

python3 main.py report

Output:

📄 Report generated: ~/.smf/optimizer-reports/report-2024-03-15.txt

View the report:

cat ~/.smf/optimizer-reports/report-2024-03-15.txt

Compare to previous report:

diff ~/.smf/optimizer-reports/report-2024-02-01.txt ~/.smf/optimizer-reports/report-2024-03-15.txt

---

6. Automating with Cron

Example: Monthly report on the 1st

0 9 1 * * python3 /home/yourname/smfworks-skills/skills/openclaw-optimizer/main.py report >> /home/yourname/logs/optimizer.log 2>&1

Example: Weekly context check on Mondays

0 8 * * 1 python3 /home/yourname/smfworks-skills/skills/openclaw-optimizer/main.py analyze --context >> /home/yourname/logs/context-check.log 2>&1

---

7. Combining with Other Skills

OpenClaw Optimizer + OpenClaw Backup: Back up before making changes based on optimizer recommendations:

python3 ~/smfworks-skills/skills/openclaw-backup/main.py
python3 ~/smfworks-skills/skills/openclaw-optimizer/main.py recommend
# Apply recommendations

OpenClaw Optimizer + Skill Manager: Remove unused skills found by the analyzer:

python3 ~/smfworks-skills/skills/openclaw-optimizer/main.py analyze --skills
python3 ~/smfworks-skills/skills/skill-manager/main.py --list
python3 ~/smfworks-skills/skills/skill-manager/main.py --remove unused-skill

---

8. Troubleshooting Common Issues

Error: SMF Works Pro subscription required

Fix: Subscribe at smfworks.com/subscribe and re-authenticate.

Audit shows high skill count but I use all of them

Fix: The threshold is a guideline. If you actively use all your skills, the warning is informational. Focus on other recommendations.

MEMORY.md keeps growing back quickly

Fix: Your agent is writing a lot to memory. Review AGENTS.md to adjust what gets captured in daily notes vs. MEMORY.md. Quality over quantity — not everything needs to be in MEMORY.md.

Large files appear in workspace that I don't recognize

Fix: Your agent or a skill may have created them. Check timestamps: ls -lat ~/.openclaw/workspace/ | head -10. If it's a backup or log file, move it out of the workspace.

---

9. Tips & Best Practices

Run monthly, not obsessively. Your workspace isn't going to bloat overnight. Monthly audits catch issues before they become real problems without adding maintenance overhead.

Back up before applying recommendations. Always run OpenClaw Backup before making significant changes to MEMORY.md or removing skills.

MEMORY.md quality beats quantity. A 5 KB MEMORY.md with the 20 most important facts about you is far more useful to your agent than a 25 KB file with 200 semi-relevant notes. Prune ruthlessly.

Treat the skill count warning as a signal, not a rule. If you have 25 skills but only use 5 regularly, consider removing the other 20. If you actively use all 25, the overhead is the cost of that capability.

Save reports for trend tracking. Keeping monthly reports lets you see if your workspace is growing faster than expected and identify what's driving the growth.