openclaw-master

name	openclaw-master
description	OpenClaw configuration expert and system doctor. Use for unclassified config questions that don't map to a more specific skill — for specific areas use their dedicated skills (see skill map below). Triggers on: openclaw expert, config doctor, system health, gateway diagnostics, openclaw master, openclaw overview, openclaw help.

Routing hints

You should route to this skill when the user has a config question that doesn't map to a more specific skill — open-ended "how do I…" questions, system doctor queries, gateway diagnostics, streaming or queue mode questions, model failover, sandbox vs tool policy, or general architecture questions. References: architecture.md, streaming.md, queue.md, model-failover.md, sandbox-vs-tool-policy-vs-elevated.md, compaction.md, session.md, agent-loop.md, system-prompt.md, soul.md, taskflow.md, oauth.md.

OpenClaw Master Reference

Comprehensive OpenClaw knowledge base. This is the top-level skill that routes to specialized skills.

Scope note: ClawDoc follows an OSS-first scope fence — internal-only modules, private runtimes, and per-host paths are never published to this repo. Each release is reviewed against this fence before tagging.

Skill map

The 23 routable OpenClaw skills, grouped by area:

Skill	When to use
Config + Gateway
openclaw-config	Config audits, gateway config keys, secrets, retry, model failover, timezone, auth profiles, authentication
openclaw-gateway	Gateway operations, HTTP/API surface (OpenAI-compatible, OpenResponses), observability (OTel, Prometheus), secrets, sandbox vs tool policy vs elevated, security audit, gateway exposure (Tailscale/runbook)
openclaw-logging	Logging configuration: log levels, log rotation, log destinations, log format, debugging log output
Memory
openclaw-memory	Memory backends (builtin/QMD/Honcho), embedding providers, active memory, dreaming, memory search
Channels
openclaw-channels	Discord, Telegram, WhatsApp, Slack, Signal, Matrix, iMessage (native imsg), routing, pairing, broadcast groups; historical BlueBubbles→iMessage migration guide at references/bluebubbles.md
Agents + Multi-agent
openclaw-agents	Multi-agent setup, bindings, sandbox, tool policies, exec approvals
Plugins
openclaw-plugins	Plugin slots, installing plugins, plugin config, context engine, plugin manifest, SDK, ClawHub registry
Tools + Workflows
openclaw-tools	Tool reference: exec, browser, cron, sessions, subagents, ACP, slash commands, Lobster, Diff viewer, Skill Workshop, media generation
openclaw-automation	Cron jobs, hooks, tasks, webhooks, polls, standing orders, Task Flow, Clawflow, Gmail Pub/Sub
openclaw-prose	OpenProse markdown-first workflow language for multi-agent AI sessions (plugin + slash command + skill pack)
openclaw-clawhub	ClawHub public registry: discover, install, publish, and verify skills and plugins; clawhub CLI
CLI + Reference
openclaw-cli	CLI commands: status, gateway, plugins, memory, agents, channels, config, sessions, cron, hooks, pairing, ACP, MCP, secrets, doctor, update, skills
openclaw-reference	General OpenClaw reference: token use, transcript hygiene, session compaction, rich output protocol, SecretRef credential surface, RELEASING
Concepts + Help
openclaw-concepts	Architecture, session, compaction, streaming, bootstrap, queue modes, agent loop, system prompt, SOUL.md, Task Flow, OAuth, presence, OpenClaw SDK
openclaw-help	Help, FAQ, "how do I..." plain-language entry point, testing
Install + Platforms + Start
openclaw-install	Install, update, migrate, uninstall. Docker, Railway, Fly, GCP, Azure, Kubernetes, Hetzner, Podman, Bun, Nix, from source
openclaw-platforms	OpenClaw on Android, iOS, macOS, Windows, Linux, Raspberry Pi, Oracle Cloud, DigitalOcean
openclaw-start	First run, bootstrapping, onboarding OpenClaw. Setup wizard, workspace bootstrap, openclaw onboard CLI
Surfaces
openclaw-nodes	Pairing and using companion nodes (iOS, Android, macOS). Camera, canvas, audio, talk, location, notifications, screen
openclaw-web	Web UI: dashboard, TUI, webchat, control UI
Providers
openclaw-providers	Model providers: OpenAI, Anthropic, Gemini, Bedrock, Ollama, DeepSeek, Groq, Together, and 60+ more
Troubleshooting
openclaw-troubleshooting	Diagnosis, error codes, ACP, hooks, pairing, channel failures, gateway runbook
Different domain — not an OpenClaw skill
clawdoc-onboarding	Guided first-run setup for ClawDoc itself (NOT OpenClaw). Walks through installing ClawDoc skills, configuring docs sync, end-to-end verification

Quick reference

Docs URLs (always current)

Topic	URL
Config reference	https://docs.openclaw.ai/gateway/configuration-reference
CLI commands	https://docs.openclaw.ai/cli/index
Channels	https://docs.openclaw.ai/channels/index
Providers	https://docs.openclaw.ai/providers/index
Tools	https://docs.openclaw.ai/tools/index
Troubleshooting	https://docs.openclaw.ai/help/troubleshooting
OpenClaw docs	https://docs.openclaw.ai
ClawHub (skills)	https://clawhub.ai

Usage quick reference

Situation	Action
Config changes	Use gateway tool (config.get / config.patch / config.apply)
Skill installation	clawhub CLI or clone to ~/.openclaw/skills/
Bot mention in Discord	<@bot-id> not plain @name
requireMention: true	Respond only when pinged
requireMention: false	Always open in that channel
Bot-to-bot messaging	Requires allowBots: "mentions" on receiving side

Config location

~/.openclaw/openclaw.json

This is the user-level config path for typical OpenClaw installs (Homebrew, pip, source). If you are running OpenClaw inside a container image that sets OPENCLAW_HOME=/data/.openclaw, the config path is /data/.openclaw/openclaw.json instead. Check openclaw doctor to see which path your install uses.

Logs

/tmp/openclaw/openclaw-YYYY-MM-DD.log

Gateway management

openclaw status              # check status
openclaw gateway restart    # restart (confirm first!)
ps aux | grep openclaw-gateway  # is it running?

Validate config

python3 -m json.tool ~/.openclaw/openclaw.json > /dev/null && echo "valid"

Key memory fix (Honcho)

plugins: { slots: { memory: "openclaw-honcho" } }

Key config paths

• agents.defaults.workspace — agent working directory
• agents.defaults.model — primary + fallback models
• agents.defaults.memorySearch — embedding provider
• memory.backend — "builtin" or "qmd"
• plugins.slots.memory — memory plugin override
• channels.discord.token — bot token
• tools.elevated.enabled — elevated exec

Common error fixes

• "plugin disabled (memory slot)" → set plugins.slots.memory
• "Port already in use" → gateway already running, don't restart
• Embedding 404 → bad memorySearch.remote.baseUrl or invalid provider
• Discord not routing → format guildId/channelId
• ACP agent not spawning → run /acp doctor, then /acp spawn codex --bind here
• Hook not firing → openclaw hooks list, then openclaw hooks enable HOOK_NAME
• Pairing code expired → codes expire after 1 hour, max 3 pending per channel

Architecture overview

Single Gateway daemon (port 18789) owns all messaging surfaces. Agents connect via WebSocket. Three memory backends: builtin (SQLite), QMD (local sidecar), Honcho (cross-session plugin). Multi-agent via agents.list[] + bindings[].

Memory architecture

Three backends:

1. Builtin — SQLite + vector search, auto-detects OpenAI/Gemini/Voyage/Mistral keys
2. QMD — Local sidecar with reranking, query expansion, extra paths
3. Honcho — Plugin openclaw-honcho, cross-session with user modeling

Tools: memory_search (semantic), memory_get (file/lines) Plus Honcho tools: honcho_context, honcho_ask, honcho_search_conclusions, honcho_search_messages, honcho_session

Active memory (optional): blocking sub-agent that runs before main reply, surfacing relevant memory proactively.

Dreaming (experimental): background consolidation with light/deep/REM phases.

Agent bootstrap files

Injected on first turn of session:

• AGENTS.md — operating instructions + memory
• SOUL.md — persona, boundaries, tone
• TOOLS.md — user-maintained tool notes
• BOOTSTRAP.md — one-time first-run ritual
• IDENTITY.md — name/emoji
• USER.md — user profile
• HEARTBEAT.md — periodic task list

Multi-agent routing

Match order (most-specific wins):

1. peer (exact DM/group/channel id)
2. parentPeer (thread inheritance)
3. guildId + roles (Discord)
4. guildId
5. teamId (Slack)
6. accountId match
7. channel-level (accountId: "*")
8. default agent

Tool profiles

Profile	Includes
minimal	session_status only
coding	group:fs + group:runtime + group:web + group:sessions + group:memory + cron + image + image_generate + video_generate
messaging	group:messaging + sessions tools
full	No restriction

Channel DM policies

Policy	Behavior
pairing (default)	Code → owner approves
allowlist	Only allowFrom list
open	Allow all (allowFrom: ["*"])
disabled	Ignore all

Gateway restart triggers

Require restart: memory.backend, plugins.entries.*.enabled, agents.defaults.*
Dynamic (no restart): tools.web.search.apiKey, agents.defaults.memorySearch.remote

CLI essentials

openclaw status          # gateway health
openclaw plugins list    # plugin status
openclaw memory status   # memory + provider
openclaw agents list     # agents + bindings
openclaw channels status # channel health
openclaw doctor          # diagnose + fix
openclaw doctor --fix    # auto-repair

Message queue modes

Mode	Behavior
collect (default)	Coalesce queued messages into single followup turn
steer	Inject immediately into current run, cancel pending tools
followup	Enqueue for next turn after current run ends
steer-backlog	Steer now AND preserve for followup

Streaming modes

• blockStreamingDefault: "on"|"off" — default for channel
• blockStreamingBreak: "text_end"|"message_end" — chunk boundary flush
• blockStreamingChunk: {minChars, maxChars} — chunk size bounds
• blockStreamingCoalesce: {minChars?, maxChars?, idleMs?} — merge blocks before send
• Discord maxLinesPerMessage (default 17) prevents UI clipping

Model failover

Two-stage: (1) auth profile rotation within provider → (2) model fallback to next in agents.defaults.model.fallbacks. Only persists fallback-owned fields to avoid overwriting manual /model changes.

ACP agents (external coding harnesses)

/acp spawn codex --bind here     # spawn and bind to current channel
/acp spawn codex --mode persistent --thread auto  # persistent
/acp status                      # check runtime state
/acp cancel                      # stop current turn
/acp close                       # close session + remove bindings
/acp doctor                      # check readiness

References structure

Each skill has a references/ directory with deep-dive docs. Key ones:

• openclaw-config/references/gateway-config.md — full config reference
• openclaw-config/references/retry.md — retry policy
• openclaw-config/references/model-failover.md — model failover + auth profiles
• openclaw-memory/references/memory-config.md — all memory config knobs
• openclaw-memory/references/memory-builtin.md — builtin engine details
• openclaw-memory/references/memory-honcho.md — Honcho plugin details
• openclaw-memory/references/dreaming.md — dreaming system
• openclaw-agents/references/multi-agent.md — full multi-agent guide
• openclaw-channels/references/discord.md — Discord channel reference
• openclaw-channels/references/channel-routing.md — routing rules + session keys
• openclaw-channels/references/pairing.md — pairing + node device pairing
• openclaw-concepts/references/architecture.md — gateway architecture
• openclaw-concepts/references/session.md — session system
• openclaw-concepts/references/compaction.md — compaction + memory flush
• openclaw-concepts/references/streaming.md — block streaming + chunking
• openclaw-concepts/references/queue.md — command queue + lane semantics
• openclaw-concepts/references/agent-loop.md — agent loop lifecycle
• openclaw-concepts/references/soul.md — SOUL.md personality guide
• openclaw-concepts/references/system-prompt.md — system prompt assembly
• openclaw-concepts/references/taskflow.md — Task Flow orchestration
• openclaw-troubleshooting/references/gateway-troubleshooting.md — deep gateway runbook
• openclaw-troubleshooting/references/hooks.md — hook system
• openclaw-plugins/references/plugin-architecture.md — plugin system
• openclaw-tools/references/acp-agents.md — ACP agent runtime
• openclaw-tools/references/lobster.md — Lobster workflow DSL
• openclaw-providers/references/providers.md — all providers

Load these for detailed reference when answering complex config questions.

Skill boundaries

• openclaw-install = how to DEPLOY OpenClaw (Docker, Railway, Fly, GCP, Azure, K8s, etc.)
• openclaw-platforms = how OpenClaw BEHAVES on specific OSes (Android, iOS, macOS, Windows, Linux, Pi, Oracle Cloud, DigitalOcean)

Use install for "how do I deploy this?" Use platforms for "what works on this OS?" Use nodes for pairing + device behavior across platforms.