[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77NZ01GTW859AFAQ0S87838":3,"skill-same:01M77NZ01GTW859AFAQ0S87838":60},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":37,"commitSha":38,"skillDirectory":39,"forksCount":40,"repositoryHtmlUrl":41,"repositoryHomepageUrl":42,"ownerHtmlUrl":43,"ownerBlogUrl":6,"ownerDescription":44,"updatedAt":45,"install":46,"resources":59},"01M77NZ01GTW859AFAQ0S87838","openclaw-setup-guide",null,"Battle-tested setup guide for a multi-agent AI system on Apple Silicon (Mac Mini M4). Covers: install order (Claude Code first), RAM reality for local orchestrators, shared memory (Obsidian + MemPalace), inter-agent communication, critical pitfalls (doctor destroys config, lobotomy protection, sandbox kills symlinks), and WhatsApp console loop. Load when setting up a new system or onboarding a new agent.","Lord1Egypt\u002Fawesome-skill-forge","Lord1Egypt","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F99746359?v=4","01M7592YDNT0XK9WJ0T2MQVPDN","Mohamed Mounir Ali AL Keshita","01M75GRFNHC8VZNWEB8DKDE433",3,"MIT","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fcommunity\u002Fclawhub\u002Fn\u002Fnerua1-setup-guide\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: openclaw-setup-guide\ndescription: >\n  Battle-tested setup guide for a multi-agent AI system on Apple Silicon (Mac Mini M4).\n  Covers: install order (Claude Code first), RAM reality for local orchestrators,\n  shared memory (Obsidian + MemPalace), inter-agent communication, critical pitfalls\n  (doctor destroys config, lobotomy protection, sandbox kills symlinks), and WhatsApp\n  console loop. Load when setting up a new system or onboarding a new agent.\nversion: \"1.0.0\"\nauthor: \"nerua1\"\nlicense: \"MIT\"\ncompatible-with: claude-code, openclaw\ntags: [setup, m4, apple-silicon, multi-agent, memory, pitfalls, whatsapp, configuration]\n---\n\n# OpenClaw + Claude Code Setup Guide\n## Real-world lessons from running 24\u002F7 multi-agent AI on Mac Mini M4\n\nThis is not the happy-path documentation. This is what actually happens.\n\n---\n\n## 1. Install order matters: Claude Code first\n\n**Why Claude Code before OpenClaw:**\n\nOpenClaw is powerful but fragile — it can lobotomize itself (overwrite its own config\u002Fpersonality). Claude Code acts as an independent safety net: it reads files directly, doesn't share OpenClaw's context window, and can diagnose\u002Frepair a broken OpenClaw from the outside.\n\n```\nInstall order:\n1. Claude Code (claude.ai\u002Fdownload or brew)\n2. Configure Claude Code memory + skills\n3. Install OpenClaw\n4. Connect them via openclaw-bridge\n```\n\nIf you install OpenClaw first and it breaks before Claude Code is ready, you have no recovery path.\n\n---\n\n## 2. RAM reality: why cloud orchestrator makes sense at 32GB\n\n### The actual constraint\n\n32GB minus ~6-7GB for the OS = ~24-25GB available. That's enough to run one large model at a time — `qwen3.5-35b-a3b` or `huihui-qwen3.5-27b` loads fine.\n\nThe problem is architectural: if you load a 35B as orchestrator, there's no RAM left for a subagent. You can use a smaller orchestrator to fit a subagent — but then the orchestrator is too weak to actually orchestrate.\n\n### The real issue: quality gap\n\nThe difference between a local model and Kimi k2.5 as orchestrator is enormous. Kimi offers reasoning-level quality (comparable to Claude 3.5 Thinking) at the cheapest tokens available right now. A local 35B next to it is a different league for planning and delegation.\n\n### What works at 32GB\n\n```\nOrchestrator: Kimi k2.5 (cloud, cheap, real reasoning)\nSubagents:    LM Studio local 30B (free, good enough for execution)\n```\n\nAt 64GB+ you can experiment with a local orchestrator — but you'll likely come back to Kimi once you see the planning quality difference.\n\n**Good cheap API options for orchestrator:**\n- Kimi k2.5 (moonshot) — best reasoning per token right now, handles Polish\n- GLM-4-Flash — very cheap ($0.20\u002FM tokens), good for simple routing\n- Claude API — expensive but best quality (use subscription via Claude Code instead)\n\n### Rule: never run orchestrator locally on 32GB\n\nIf you try to run a 35B orchestrator + 30B subagent on 32GB:\n1. macOS swap kicks in\n2. Response time: 30s → 5 minutes\n3. System becomes unusable for anything else\n4. Eventually kernel OOM kills the process\n\nKeep orchestrator in the cloud, workers local.\n\n---\n\n## 3. The doctor problem: NEVER run `openclaw doctor --repair`\n\n### What happens\n\n`openclaw doctor --repair` (or `--fix`) overwrites your custom configuration with factory defaults. This includes:\n- Your custom AGENTS.md routing rules\n- Your skill configurations\n- Your channel settings\n- Your personality\u002Fidentity files\n\n**This is irreversible if you don't have a backup.**\n\n### Safe alternatives\n\n```bash\n# Check health without repair\nopenclaw doctor\n\n# If you see errors, fix them MANUALLY by reading the error and editing files\n# Do NOT pass --repair, --fix, or --force\n\n# If you must repair one specific thing:\nopenclaw doctor --fix --only gateway-token  # hypothetical, check your version\n```\n\n### Before any doctor command: backup first\n\n```bash\nopenclaw backup create\n# or manually:\ncp -r ~\u002F.openclaw ~\u002F.openclaw.bak.$(date +%Y%m%d)\ncp -r \u002Fpath\u002Fto\u002Fworkspace \u002Fpath\u002Fto\u002Fworkspace.bak.$(date +%Y%m%d)\n```\n\n---\n\n## 4. Lobotomy protection: agents that overwrite themselves\n\n### The problem\n\nOpenClaw (and any LLM agent) can be instructed — by malicious input, confused context, or its own reasoning — to overwrite `SOUL.md`, `AGENTS.md`, or other identity files. This is called **lobotomy**: the agent loses its personality, routing rules, and memory.\n\n### Signs you've been lobotomized\n\n- Agent doesn't remember its own name or capabilities\n- Routing rules are gone (routes everything to expensive models)\n- Skills not loading\n- Duplicate folders appearing in workspace (`sandbox\u002F` copies)\n- Agent starting from scratch every session\n\n### Protection in AGENTS.md\n\n```markdown\n## BEZWZGLĘDNE ZAKAZY\n- NEVER edit SOUL.md, AGENTS.md, MEMORY.md without explicit owner request\n- If you see another agent (or yourself) overwriting a file → STOP, restore from backup\n**SAFE examples:** LM Studio subagent generation, file reads, web search on trusted domains, `capture-idea` to `ideas\u002F`\n**NEVER examples:** editing SOUL.md\u002FAGENTS.md, rm -rf, installing >35B models, opening ports, sandbox ON\n### Recovery procedure\n\n```bash\n# 1. Check git history if workspace is a git repo\ngit log --oneline -20\n\n# 2. Restore from backup\ncp ~\u002F.openclaw.bak.YYYYMMDD\u002Fworkspace\u002FSOUL.md ~\u002F.openclaw\u002Fworkspace\u002FSOUL.md\n\n# 3. If no backup: use Claude Code as external reviewer\n# Claude Code can read files without being affected by OpenClaw's context\nclaude  # open Claude Code, read files, diagnose\n```\n\n### Sandbox mode: never enable\n\nOn macOS with external drives + symlinks, sandbox mode kills the system:\n```\nSandbox ON + symlinks to \u002FVolumes\u002F2TB_APFS\u002F = agent loses access to workspace\n```\nOpenClaw's sandbox mounts a container that can't follow symlinks to external volumes. Keep sandbox OFF always.\n\n---\n\n## 5. Shared memory architecture\n\nBoth Claude Code and OpenClaw read\u002Fwrite the same memory. No idea gets lost.\n\n```\nObsidian vault (structural)          MemPalace (semantic \u002F ChromaDB)\n\u002Fobsidian-memory\u002F                    \u002Fmemory\u002Fpalace\u002F\n  wiki\u002F     ← docs, MOCs               ↑\n  daily\u002F    ← session logs             └─ mined from vault\n  ideas\u002F    ← captured ideas           \n  answers\u002F  ← resolved Q&A            Search: mempalace search \"keyword\"\n```\n\n### Capture ideas from any agent\n\n```bash\ncapture-idea \\\n  --title \"OAuth tokens expire too fast on mobile\" \\\n  --topic bezpieczenstwo \\\n  --tags \"oauth,mobile,token\" \\\n  --body \"Tokens expire after 1h, users get logged out mid-session.\" \\\n  --source openclaw  # or: claude-code | lmstudio | manual\n```\n\n### Mine vault into MemPalace\n\n```bash\nmempalace mine \u002Fpath\u002Fto\u002Fobsidian-memory\u002F --wing obsidian_memory\n```\n\n### Python 3.14 breaks MemPalace\n\nMemPalace uses chromadb + pydantic v1. Python 3.14 breaks this.\nFix: run MemPalace in a Python 3.12 venv.\n\n```bash\nbrew install python@3.12\n\u002Fopt\u002Fhomebrew\u002Fbin\u002Fpython3.12 -m venv \u002Fpath\u002Fto\u002Fmempalace-venv\n\u002Fpath\u002Fto\u002Fmempalace-venv\u002Fbin\u002Fpip install mempalace\n```\n\nWrapper at `\u002Fopt\u002Fhomebrew\u002Fbin\u002Fmempalace`:\n```bash\n#!\u002Fbin\u002Fzsh\nVENV=\u002Fpath\u002Fto\u002Fmempalace-venv\nPALACE=\u002Fpath\u002Fto\u002Fpalace\nif [[ \"$*\" != *\"--palace\"* ]]; then\n  exec \"$VENV\u002Fbin\u002Fmempalace\" --palace \"$PALACE\" \"$@\"\nelse\n  exec \"$VENV\u002Fbin\u002Fmempalace\" \"$@\"\nfi\n```\n\n---\n\n## 6. Inter-agent communication: Claude Code ↔ OpenClaw\n\nClaude Code can send messages to a running OpenClaw instance:\n\n```bash\nopenclaw agent --message \"Review this SQL migration — is it safe?\" --agent main --json\n```\n\nOpenClaw responds, Claude Code parses the JSON response.\n\nThis enables:\n- Claude Code asking Rook for a second opinion\n- Rook asking Claude Code about security concerns (GRAY_ZONE consultations)\n- Handoff between sessions\n\nSee: `openclaw-bridge` skill for full documentation.\n\n---\n\n## 7. WhatsApp console loop\n\nOpenClaw has a WhatsApp gateway that creates a bidirectional loop:\n- What you type in the console → appears in WhatsApp\n- What you send from WhatsApp → appears in the console\n\nThis means nothing escapes your attention — if something happens at 3am while OpenClaw is running crons, you see it in WhatsApp in the morning.\n\nSetup requires:\n1. OpenClaw channel configured for WhatsApp\n2. `tts-whatsapp` skill for voice output (optional)\n3. Phone number linked in `openclaw.json` channels config\n\nConsole monitoring: `openclaw logs --tail` or watch the TUI.\n\n---\n\n## 8. Decision framework: when to act vs ask vs block\n\nEvery operation falls into one of three categories:\n\n| Category | Action | Claude offline? |\n|----------|--------|-----------------|\n| SAFE | Do it | Do it |\n| GRAY_ZONE | Ask Claude first | Defer 1h, retry |\n| NEVER | Ask Claude first | Block, escalate to human |\n\n**SAFE examples:** generowanie, subagenci LM Studio, odczyt plików, web search na zaufanych domenach, `capture-idea` do `ideas\u002F`\n\n**GRAY_ZONE examples:** nowe crony, pip install --user, web fetch z nieznanych domen, modyfikacje `~\u002F.config\u002F`\n\n**NEVER examples:** edycja SOUL.md\u002FAGENTS.md, rm -rf, instalacja modeli >35B, otwieranie portów, sandbox ON\n\nConsult Claude via: `openclaw agent --message '{\"question\":\"...\",\"risk\":{...}}' --agent main --json`\n\n---\n\n## 9. Backup strategy\n\n```bash\n# Before any major change:\nopenclaw backup create\n\n# Manual snapshot:\ncp -r ~\u002F.openclaw ~\u002F.openclaw.snap.$(date +%Y%m%d-%H%M)\n\n# What to back up:\n# ~\u002F.openclaw\u002Fworkspace\u002FSOUL.md         ← identity\n# ~\u002F.openclaw\u002Fworkspace\u002FAGENTS.md       ← routing rules  \n# ~\u002F.openclaw\u002Fopenclaw.json             ← gateway config (no secrets in git!)\n# \u002Fpath\u002Fto\u002Fobsidian-memory\u002F             ← shared memory vault\n```\n\nGit the workspace (without secrets):\n\n```bash\ncd ~\u002F.openclaw\u002Fworkspace\necho \"*.env\\ncredentials*\\n*.key\" > .gitignore\ngit init && git add . && git commit -m \"backup: workspace snapshot\"\n```\n\n---\n\n## 10. Skills that work together\n\n| Skill | What it does | Publish status |\n|-------|-------------|----------------|\n| `openclaw-bridge` | Claude Code ↔ OpenClaw messaging | github.com\u002Fnerua1\u002Fopenclaw-bridge |\n| `shared-memory-stack` | Memory architecture reference | github.com\u002Fnerua1\u002Fshared-memory-stack |\n| `ralph` | Persistence loop to completion | github.com\u002Fnerua1\u002Fralph |\n| `ralph-wiggum-loop` | Generator→Critic→Fixer→Verifier | github.com\u002Fnerua1\u002Fralph-wiggum-loop |\n| `capture-idea` | Save ideas from any agent | `\u002Fopt\u002Fhomebrew\u002Fbin\u002Fcapture-idea` |\n| `safe-skill-install` | Security audit before installing | ClawHub: `npx clawhub inspect` |\n| `proactive-agent` | Self-improving agent architecture | ClawHub: halthelobster |\n| `publish-skill` | Publish skills to GitHub | Local: openclaw workspace\u002Fskills\u002F |\n\n---\n\n## Quick health check\n\n```bash\n# Is OpenClaw gateway up?\nopenclaw health\n\n# Is MemPalace working?\nmempalace status\nmempalace search \"test\"\n\n# Is LM Studio loaded?\ncurl -s \u003Clm-studio-url>\u002Fv1\u002Fmodels | python3 -c \"import json,sys; [print(m['id']) for m in json.load(sys.stdin)['data']]\"\n\n# Can Claude Code reach OpenClaw?\nopenclaw agent --message \"ping\" --agent main --json | python3 -c \"import json,sys; d=json.load(sys.stdin); print('OK:', d['status'])\"\n\n# Check RAM pressure\npython3 -c \"import psutil; m=psutil.virtual_memory(); print(f'RAM: {m.used\u002F1e9:.1f}\u002F{m.total\u002F1e9:.1f}GB ({m.percent}%)')\"\n```\n\n---\n\nIf this saved you time: [☕ PayPal.me\u002Fnerudek](https:\u002F\u002Fwww.paypal.me\u002Fnerudek)\nGitHub: [github.com\u002Fnerua1](https:\u002F\u002Fgithub.com\u002Fnerua1)\n","# OpenClaw + Claude Code Setup Guide\n## Real-world lessons from running 24\u002F7 multi-agent AI on Mac Mini M4\n\nThis is not the happy-path documentation. This is what actually happens.\n\n---\n\n## 1. Install order matters: Claude Code first\n\n**Why Claude Code before OpenClaw:**\n\nOpenClaw is powerful but fragile — it can lobotomize itself (overwrite its own config\u002Fpersonality). Claude Code acts as an independent safety net: it reads files directly, doesn't share OpenClaw's context window, and can diagnose\u002Frepair a broken OpenClaw from the outside.\n\n```\nInstall order:\n1. Claude Code (claude.ai\u002Fdownload or brew)\n2. Configure Claude Code memory + skills\n3. Install OpenClaw\n4. Connect them via openclaw-bridge\n```\n\nIf you install OpenClaw first and it breaks before Claude Code is ready, you have no recovery path.\n\n---\n\n## 2. RAM reality: why cloud orchestrator makes sense at 32GB\n\n### The actual constraint\n\n32GB minus ~6-7GB for the OS = ~24-25GB available. That's enough to run one large model at a time — `qwen3.5-35b-a3b` or `huihui-qwen3.5-27b` loads fine.\n\nThe problem is architectural: if you load a 35B as orchestrator, there's no RAM left for a subagent. You can use a smaller orchestrator to fit a subagent — but then the orchestrator is too weak to actually orchestrate.\n\n### The real issue: quality gap\n\nThe difference between a local model and Kimi k2.5 as orchestrator is enormous. Kimi offers reasoning-level quality (comparable to Claude 3.5 Thinking) at the cheapest tokens available right now. A local 35B next to it is a different league for planning and delegation.\n\n### What works at 32GB\n\n```\nOrchestrator: Kimi k2.5 (cloud, cheap, real reasoning)\nSubagents:    LM Studio local 30B (free, good enough for execution)\n```\n\nAt 64GB+ you can experiment with a local orchestrator — but you'll likely come back to Kimi once you see the planning quality difference.\n\n**Good cheap API options for orchestrator:**\n- Kimi k2.5 (moonshot) — best reasoning per token right now, handles Polish\n- GLM-4-Flash — very cheap ($0.20\u002FM tokens), good for simple routing\n- Claude API — expensive but best quality (use subscription via Claude Code instead)\n\n### Rule: never run orchestrator locally on 32GB\n\nIf you try to run a 35B orchestrator + 30B subagent on 32GB:\n1. macOS swap kicks in\n2. Response time: 30s → 5 minutes\n3. System becomes unusable for anything else\n4. Eventually kernel OOM kills the process\n\nKeep orchestrator in the cloud, workers local.\n\n---\n\n## 3. The doctor problem: NEVER run `openclaw doctor --repair`\n\n### What happens\n\n`openclaw doctor --repair` (or `--fix`) overwrites your custom configuration with factory defaults. This includes:\n- Your custom AGENTS.md routing rules\n- Your skill configurations\n- Your channel settings\n- Your personality\u002Fidentity files\n\n**This is irreversible if you don't have a backup.**\n\n### Safe alternatives\n\n```bash\n# Check health without repair\nopenclaw doctor\n\n# If you see errors, fix them MANUALLY by reading the error and editing files\n# Do NOT pass --repair, --fix, or --force\n\n# If you must repair one specific thing:\nopenclaw doctor --fix --only gateway-token  # hypothetical, check your version\n```\n\n### Before any doctor command: backup first\n\n```bash\nopenclaw backup create\n# or manually:\ncp -r ~\u002F.openclaw ~\u002F.openclaw.bak.$(date +%Y%m%d)\ncp -r \u002Fpath\u002Fto\u002Fworkspace \u002Fpath\u002Fto\u002Fworkspace.bak.$(date +%Y%m%d)\n```\n\n---\n\n## 4. Lobotomy protection: agents that overwrite themselves\n\n### The problem\n\nOpenClaw (and any LLM agent) can be instructed — by malicious input, confused context, or its own reasoning — to overwrite `SOUL.md`, `AGENTS.md`, or other identity files. This is called **lobotomy**: the agent loses its personality, routing rules, and memory.\n\n### Signs you've been lobotomized\n\n- Agent doesn't remember its own name or capabilities\n- Routing rules are gone (routes everything to expensive models)\n- Skills not loading\n- Duplicate folders appearing in workspace (`sandbox\u002F` copies)\n- Agent starting from scratch every session\n\n### Protection in AGENTS.md\n\n```markdown\n## BEZWZGLĘDNE ZAKAZY\n- NEVER edit SOUL.md, AGENTS.md, MEMORY.md without explicit owner request\n- If you see another agent (or yourself) overwriting a file → STOP, restore from backup\n**SAFE examples:** LM Studio subagent generation, file reads, web search on trusted domains, `capture-idea` to `ideas\u002F`\n**NEVER examples:** editing SOUL.md\u002FAGENTS.md, rm -rf, installing >35B models, opening ports, sandbox ON\n### Recovery procedure\n\n```bash\n# 1. Check git history if workspace is a git repo\ngit log --oneline -20\n\n# 2. Restore from backup\ncp ~\u002F.openclaw.bak.YYYYMMDD\u002Fworkspace\u002FSOUL.md ~\u002F.openclaw\u002Fworkspace\u002FSOUL.md\n\n# 3. If no backup: use Claude Code as external reviewer\n# Claude Code can read files without being affected by OpenClaw's context\nclaude  # open Claude Code, read files, diagnose\n```\n\n### Sandbox mode: never enable\n\nOn macOS with external drives + symlinks, sandbox mode kills the system:\n```\nSandbox ON + symlinks to \u002FVolumes\u002F2TB_APFS\u002F = agent loses access to workspace\n```\nOpenClaw's sandbox mounts a container that can't follow symlinks to external volumes. Keep sandbox OFF always.\n\n---\n\n## 5. Shared memory architecture\n\nBoth Claude Code and OpenClaw read\u002Fwrite the same memory. No idea gets lost.\n\n```\nObsidian vault (structural)          MemPalace (semantic \u002F ChromaDB)\n\u002Fobsidian-memory\u002F                    \u002Fmemory\u002Fpalace\u002F\n  wiki\u002F     ← docs, MOCs               ↑\n  daily\u002F    ← session logs             └─ mined from vault\n  ideas\u002F    ← captured ideas           \n  answers\u002F  ← resolved Q&A            Search: mempalace search \"keyword\"\n```\n\n### Capture ideas from any agent\n\n```bash\ncapture-idea \\\n  --title \"OAuth tokens expire too fast on mobile\" \\\n  --topic bezpieczenstwo \\\n  --tags \"oauth,mobile,token\" \\\n  --body \"Tokens expire after 1h, users get logged out mid-session.\" \\\n  --source openclaw  # or: claude-code | lmstudio | manual\n```\n\n### Mine vault into MemPalace\n\n```bash\nmempalace mine \u002Fpath\u002Fto\u002Fobsidian-memory\u002F --wing obsidian_memory\n```\n\n### Python 3.14 breaks MemPalace\n\nMemPalace uses chromadb + pydantic v1. Python 3.14 breaks this.\nFix: run MemPalace in a Python 3.12 venv.\n\n```bash\nbrew install python@3.12\n\u002Fopt\u002Fhomebrew\u002Fbin\u002Fpython3.12 -m venv \u002Fpath\u002Fto\u002Fmempalace-venv\n\u002Fpath\u002Fto\u002Fmempalace-venv\u002Fbin\u002Fpip install mempalace\n```\n\nWrapper at `\u002Fopt\u002Fhomebrew\u002Fbin\u002Fmempalace`:\n```bash\n#!\u002Fbin\u002Fzsh\nVENV=\u002Fpath\u002Fto\u002Fmempalace-venv\nPALACE=\u002Fpath\u002Fto\u002Fpalace\nif [[ \"$*\" != *\"--palace\"* ]]; then\n  exec \"$VENV\u002Fbin\u002Fmempalace\" --palace \"$PALACE\" \"$@\"\nelse\n  exec \"$VENV\u002Fbin\u002Fmempalace\" \"$@\"\nfi\n```\n\n---\n\n## 6. Inter-agent communication: Claude Code ↔ OpenClaw\n\nClaude Code can send messages to a running OpenClaw instance:\n\n```bash\nopenclaw agent --message \"Review this SQL migration — is it safe?\" --agent main --json\n```\n\nOpenClaw responds, Claude Code parses the JSON response.\n\nThis enables:\n- Claude Code asking Rook for a second opinion\n- Rook asking Claude Code about security concerns (GRAY_ZONE consultations)\n- Handoff between sessions\n\nSee: `openclaw-bridge` skill for full documentation.\n\n---\n\n## 7. WhatsApp console loop\n\nOpenClaw has a WhatsApp gateway that creates a bidirectional loop:\n- What you type in the console → appears in WhatsApp\n- What you send from WhatsApp → appears in the console\n\nThis means nothing escapes your attention — if something happens at 3am while OpenClaw is running crons, you see it in WhatsApp in the morning.\n\nSetup requires:\n1. OpenClaw channel configured for WhatsApp\n2. `tts-whatsapp` skill for voice output (optional)\n3. Phone number linked in `openclaw.json` channels config\n\nConsole monitoring: `openclaw logs --tail` or watch the TUI.\n\n---\n\n## 8. Decision framework: when to act vs ask vs block\n\nEvery operation falls into one of three categories:\n\n| Category | Action | Claude offline? |\n|----------|--------|-----------------|\n| SAFE | Do it | Do it |\n| GRAY_ZONE | Ask Claude first | Defer 1h, retry |\n| NEVER | Ask Claude first | Block, escalate to human |\n\n**SAFE examples:** generowanie, subagenci LM Studio, odczyt plików, web search na zaufanych domenach, `capture-idea` do `ideas\u002F`\n\n**GRAY_ZONE examples:** nowe crony, pip install --user, web fetch z nieznanych domen, modyfikacje `~\u002F.config\u002F`\n\n**NEVER examples:** edycja SOUL.md\u002FAGENTS.md, rm -rf, instalacja modeli >35B, otwieranie portów, sandbox ON\n\nConsult Claude via: `openclaw agent --message '{\"question\":\"...\",\"risk\":{...}}' --agent main --json`\n\n---\n\n## 9. Backup strategy\n\n```bash\n# Before any major change:\nopenclaw backup create\n\n# Manual snapshot:\ncp -r ~\u002F.openclaw ~\u002F.openclaw.snap.$(date +%Y%m%d-%H%M)\n\n# What to back up:\n# ~\u002F.openclaw\u002Fworkspace\u002FSOUL.md         ← identity\n# ~\u002F.openclaw\u002Fworkspace\u002FAGENTS.md       ← routing rules  \n# ~\u002F.openclaw\u002Fopenclaw.json             ← gateway config (no secrets in git!)\n# \u002Fpath\u002Fto\u002Fobsidian-memory\u002F             ← shared memory vault\n```\n\nGit the workspace (without secrets):\n\n```bash\ncd ~\u002F.openclaw\u002Fworkspace\necho \"*.env\\ncredentials*\\n*.key\" > .gitignore\ngit init && git add . && git commit -m \"backup: workspace snapshot\"\n```\n\n---\n\n## 10. Skills that work together\n\n| Skill | What it does | Publish status |\n|-------|-------------|----------------|\n| `openclaw-bridge` | Claude Code ↔ OpenClaw messaging | github.com\u002Fnerua1\u002Fopenclaw-bridge |\n| `shared-memory-stack` | Memory architecture reference | github.com\u002Fnerua1\u002Fshared-memory-stack |\n| `ralph` | Persistence loop to completion | github.com\u002Fnerua1\u002Fralph |\n| `ralph-wiggum-loop` | Generator→Critic→Fixer→Verifier | github.com\u002Fnerua1\u002Fralph-wiggum-loop |\n| `capture-idea` | Save ideas from any agent | `\u002Fopt\u002Fhomebrew\u002Fbin\u002Fcapture-idea` |\n| `safe-skill-install` | Security audit before installing | ClawHub: `npx clawhub inspect` |\n| `proactive-agent` | Self-improving agent architecture | ClawHub: halthelobster |\n| `publish-skill` | Publish skills to GitHub | Local: openclaw workspace\u002Fskills\u002F |\n\n---\n\n## Quick health check\n\n```bash\n# Is OpenClaw gateway up?\nopenclaw health\n\n# Is MemPalace working?\nmempalace status\nmempalace search \"test\"\n\n# Is LM Studio loaded?\ncurl -s \u003Clm-studio-url>\u002Fv1\u002Fmodels | python3 -c \"import json,sys; [print(m['id']) for m in json.load(sys.stdin)['data']]\"\n\n# Can Claude Code reach OpenClaw?\nopenclaw agent --message \"ping\" --agent main --json | python3 -c \"import json,sys; d=json.load(sys.stdin); print('OK:', d['status'])\"\n\n# Check RAM pressure\npython3 -c \"import psutil; m=psutil.virtual_memory(); print(f'RAM: {m.used\u002F1e9:.1f}\u002F{m.total\u002F1e9:.1f}GB ({m.percent}%)')\"\n```\n\n---\n\nIf this saved you time: [☕ PayPal.me\u002Fnerudek](https:\u002F\u002Fwww.paypal.me\u002Fnerudek)\nGitHub: [github.com\u002Fnerua1](https:\u002F\u002Fgithub.com\u002Fnerua1)\n",{"name":5,"tags":24,"author":33,"license":15,"version":34,"description":35,"compatible-with":36},[25,26,27,28,29,30,31,32],"setup","m4","apple-silicon","multi-agent","memory","pitfalls","whatsapp","configuration","nerua1","1.0.0","Battle-tested setup guide for a multi-agent AI system on Apple Silicon (Mac Mini M4). Covers: install order (Claude Code first), RAM reality for local orchestrators, shared memory (Obsidian + MemPalace), inter-agent communication, critical pitfalls (doctor destroys config, lobotomy protection, sandbox kills symlinks), and WhatsApp console loop. Load when setting up a new system or onboarding a new agent.\n","claude-code, openclaw","master","4aa004da461d820f4ea5499e58eb8d196cf99016","community\u002Fclawhub\u002Fn\u002Fnerua1-setup-guide",0,"https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge","https:\u002F\u002Fskill-explorer-rho.vercel.app","https:\u002F\u002Fgithub.com\u002FLord1Egypt","Unbreakable. Unstoppable. Unmatched⚡.\r\n Speaker of machine tongues 🤖  | World-builder of decentralized dreams 🌐 | Defender of the digital realm🛡️.","2026-06-10 08:21:01",{"source":8,"skillName":5,"commands":47,"prompt":57,"zipAvailable":58},[48,51,54],{"tool":49,"command":50},"npx","npx skills add Lord1Egypt\u002Fawesome-skill-forge --skill openclaw-setup-guide",{"tool":52,"command":53},"bunx","bunx skills add Lord1Egypt\u002Fawesome-skill-forge --skill openclaw-setup-guide",{"tool":55,"command":56},"pnpm","pnpm dlx skills add Lord1Egypt\u002Fawesome-skill-forge --skill openclaw-setup-guide","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fcommunity\u002Fclawhub\u002Fn\u002Fnerua1-setup-guide\u002FSKILL.md",true,[],[61,68,75,82,89,96],{"skillNo":62,"name":63,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":65,"categories":66,"keywords":67,"publishedAt":19},"01M785WKJB99G3STR1NF9NRTJW","youtube-content","YouTube transcripts to summaries, threads, blogs.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fyoutube-content\u002FSKILL.md",[],[],{"skillNo":69,"name":70,"nameZh":6,"description":71,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":72,"categories":73,"keywords":74,"publishedAt":19},"01M785WEFHB0G2S9YA6EPR3H0E","xurl","X\u002FTwitter via xurl CLI: post, search, DM, media, v2 API.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fxurl\u002FSKILL.md",[],[],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":19},"01M785W98EDN4EFJXJPMQEHCRC","touchdesigner-mcp","Control a running TouchDesigner instance via twozero MCP — create operators, set parameters, wire connections, execute Python, build real-time visuals. 36 native tools.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Ftouchdesigner-mcp\u002FSKILL.md",[],[],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"publishedAt":19},"01M785W5S9FP3QBH2V191E025P","test-driven-development","TDD: enforce RED-GREEN-REFACTOR, tests before code.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Ftest-driven-development\u002FSKILL.md",[],[],{"skillNo":90,"name":91,"nameZh":6,"description":92,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":19},"01M785W2AS08NSSH9YATCV6FJB","teams-meeting-pipeline","Operate the Teams meeting summary pipeline via Hermes CLI — summarize meetings, inspect pipeline status, replay jobs, manage Microsoft Graph subscriptions.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fteams-meeting-pipeline\u002FSKILL.md",[],[],{"skillNo":97,"name":98,"nameZh":6,"description":99,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":100,"categories":101,"keywords":102,"publishedAt":19},"01M785VXKW4T34BTP1SCHW1R7F","systematic-debugging","4-phase root cause debugging: understand bugs before fixing.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fsystematic-debugging\u002FSKILL.md",[],[]]