OpenClaw to NanoClaw Migration

Migrate from OpenClaw

Guide the user through migrating their OpenClaw installation into NanoClaw v2. This is a conversation, not a batch job. Read OpenClaw state, discuss it with the user, decide together what to bring over and where it belongs in v2's entity model, and show proposed changes before applying.

Principle: Never silently copy data. Read it, explain it, place it, then apply. Credentials are masked when displayed (first 4 + ... + last 4). Make judgment calls about what's core vs. reference material.

UX: Use AskUserQuestion for multiple-choice only. Use plain text for free-form input. Don't dump raw data — summarize and explain conversationally.

What this skill changes (conformance)

This skill drives existing NanoClaw entry points (setup/index.ts --step register, scripts/init-first-agent.ts, the onecli CLI) and copies a few files in (workspace markdown, OpenClaw skills, and its own transform module + test). It makes no code-level reach-in into core. Its integration assumptions about v2 are guarded by scripts/transform.test.ts, which is copied into the project's scripts/ test tree on apply (Phase 8) so vitest runs it against the composed install. REMOVE.md reverses every file the skill copies.

v2 architecture the migration targets

OpenClaw and NanoClaw v2 differ structurally. Keep these in mind throughout:

• Entity model. v2's central DB (data/v2.db) holds users, user_roles, agent_groups, messaging_groups, and the messaging_group_agents wiring between them. There is no store/messages.db and no scheduled_tasks table.
• Container isolation. Each agent group runs in its own Linux container. An OpenClaw "agent" maps to a v2 agent group (workspace + memory + CLAUDE.md); an OpenClaw chat/group maps to a v2 messaging group; the wiring row connects them.
• Standing instructions vs memory. Per-group role, personality, and behavior live in groups/<folder>/instructions.prepend.md. Durable facts live under groups/<folder>/memory/. The provider project document is composed at spawn and must not be edited.
• Credentials. Container-facing API credentials (Anthropic, OpenAI, …) are held in the OneCLI Agent Vault and injected per request — never in container env vars. Host-side channel tokens (Telegram/Discord/Slack bot tokens) stay in .env; the NanoClaw host process reads them to connect to the platform.
• Access control. Per messaging group unknown_sender_policy plus user_roles (owner/admin) and agent_group_members — not a JSON allowlist file.
• Scheduled tasks. A task is a messages_in row (kind='task') in a session's inbound.db, carrying a cron recurrence and a process_after timestamp. The agent creates them via its schedule_task MCP tool.

Migration State File

Create migration-state.md in the project root at the start of Phase 0. Update it after each phase. It's the single source of truth — if context is lost, re-read it to recover decisions and progress. Re-read it before starting any phase.

Sections to maintain:

• Progress — checkbox list of phases (Phase 0–8)
• Discovery — STATE_DIR, IDENTITY_NAME, channels, groups (with v2 platform_id mappings), workspace files, cron count, MCP servers
• Decisions — assistant_name, shared-vs-separate, primary owner agent
• Owner & Primary Agent — user id, role, agent group folder
• Registered Groups — table: folder, platform_id, channel, session_mode
• Credentials — table: credential, destination (vault / .env), status
• Settings Migrated — timezone, container timeout
• Identity & Memory — prepend and memory paths created for each group
• Scheduled Tasks — table: original_id, name, mapped schedule, status
• Deferred / Not Applicable — unsupported channels, OpenClaw-only features

Keep it factual and terse. Delete it at the end of Phase 8 (or offer to keep it as a record).

Phase 0: Discovery

Run the discovery script to find and summarize the OpenClaw installation:

pnpm exec tsx ${CLAUDE_SKILL_DIR}/scripts/discover-openclaw.ts

If the user specifies a custom path, pass --state-dir <path>.

Parse the status block. Key fields: STATUS, STATE_DIR, CHANNELS, WORKSPACE_FILES, DAILY_MEMORY_FILES, SKILL_COUNT, SKILLS, CRON_JOBS, MCP_SERVERS, IDENTITY_NAME, AGENT_COUNT, AGENT_IDS, GROUPS (each formatted channel:id(name)=>v2_platform_id — the right-hand value is what to pass as --platform-id to register).

Sanity-check the output. The script detects known structures but can miss data if OpenClaw's format changed. Check CONFIG_TOP_KEYS and CONFIG_CHANNEL_KEYS — if you see keys it didn't report on, read that section of the config with the Read tool. Check STATE_DIR_CONTENTS for directories it doesn't scan.

If STATUS=not_found: Tell the user no OpenClaw install was detected at the standard locations (~/.openclaw, ~/.clawdbot). Ask for a custom path; if none, exit.

If STATUS=found: Present a human-readable summary (identity name, workspace files, channels and which v2 supports, daily memory count, skills, cron count, MCP servers, agent count). Then paraphrase the key architectural differences from the section above — don't dump it as a table.

AskUserQuestion: "Ready to start migrating? I'll go through each area one at a time."

1. Yes, let's go — proceed to Phase 1
2. Tell me more — explain any area they ask about
3. Skip migration — exit

Phase 1: Agents, Groups, and Shared vs Separate

Decide this before identity/memory — it determines where files go.

OpenClaw model: all groups routed to one agent share a workspace (SOUL/MEMORY/IDENTITY) and personality; only the session is per-group.

v2 model: each agent group is a separate container with its own filesystem, standing instructions, and memory/ tree. Multiple messaging groups wired to the same agent group share that state. There is no groups/global/.

AskUserQuestion: "In OpenClaw your groups shared one personality and memory. In v2 each agent group is separate. How do you want to handle this?"

1. Shared identity (recommended if it was one bot) — apply the same core identity to each selected group's instructions.prepend.md; keep group facts in each group's memory tree.
2. Fully separate — each group gets independent memory and instructions; no shared base edit.
3. Just the primary agent for now — set one agent up; add others later.

Remember this choice for Phase 3.

Confirm the assistant name

IDENTITY_NAME from discovery is the OpenClaw name. Ask: "Your OpenClaw assistant was named <IDENTITY_NAME>. Keep it in v2?" If empty, ask them to choose (default: "Andy"). The chosen name is passed as --assistant-name to register/init.

Seed the owner and the primary DM agent

The owner identity and the primary agent are created together by scripts/init-first-agent.ts. It upserts the user, grants the owner role, creates the agent group + filesystem, wires a DM messaging group, and queues a welcome DM over the running service's CLI socket — so the service must be running. If it isn't, tell the user to start it first.

Resolve the owner's channel identity and the DM platform id (use the channel's own terminology). Then:

pnpm exec tsx scripts/init-first-agent.ts \
  --channel <channel> \
  --user-id <channel>:<handle> \
  --platform-id <channel>:<dm-id> \
  --display-name "<Owner Name>" \
  --agent-name "<confirmed assistant name>" \
  [--role owner]      # default: owner

For direct-addressable channels (telegram, whatsapp) the --platform-id is usually the same handle as --user-id with the channel prefix. --role defaults to owner (global, cross-channel) — use admin (scoped to the agent group) or member only if intended.

Register the remaining groups

For each additional OpenClaw group the user wants to bring over, register a messaging group and wire it to an agent group:

pnpm exec tsx setup/index.ts --step register -- \
  --platform-id "<v2_platform_id from discovery>" \
  --name "<group name>" \
  --folder "<channel>_<name-slug>" \
  --channel "<channel>" \
  --session-mode "<shared|agent-shared|per-thread>" \
  [--trigger "@<assistant name>"] \
  [--no-trigger-required] \
  --assistant-name "<assistant name>"

Notes:

• register namespaces the --platform-id the same way the adapter will at runtime, so pass the => value discovery emitted (or the raw OpenClaw id).
• Reuse a --folder to put a group on an existing agent (shared base/separate conversations); use a new --folder for a fully separate agent.
• Engage defaults come from the channel adapter's declaration (most group chats default to mention-based engagement; channels without a mention signal default to a name pattern). Pass --trigger to set an explicit regex, or --no-trigger-required for respond-to-everything.
• Register groups from channels v2 doesn't support yet too — the messaging group and wiring persist and activate when that channel is installed.

Folder naming: <channel>_<name-slug> (e.g. telegram_dev-team). Confirm each name and folder with the user.

Phase 2: Settings from Config

Read the config (<STATE_DIR>/openclaw.json or clawdbot.json) for settings that map to v2 setup.

Timezone

Check agents.defaults.userTimezone. If it's a valid IANA zone, write it to .env as TZ=<timezone>. v2 reads TZ from .env (src/config.ts) and uses it for cron/recurrence evaluation, so this matters for scheduled tasks.

Container timeout

Check agents.defaults.timeoutSeconds. v2's equivalent is CONTAINER_TIMEOUT (env var, default 30 min) or per-group ncl groups config update. If the OpenClaw value differs notably, note it; the user can set CONTAINER_TIMEOUT=<ms> in .env.

Access control (sender policies)

OpenClaw per-channel allowFrom / dmPolicy / groupPolicy map onto v2's model, which is not a JSON file. Each messaging group has an unknown_sender_policy; access is granted via user_roles (owner/admin) and agent_group_members. Map:

• dmPolicy/groupPolicy: "open" → leave the default; no extra grants.

• allowFrom / groupAllowFrom lists → for each allowed sender, upsert the user and add them as a member of the relevant agent group via ncl:

  ncl users create --id "<channel>:<handle>" --kind <channel> --display-name "<name>"
  ncl members add --user "<channel>:<handle>" --group "<ag-id>"

• dmPolicy: "disabled" → don't wire that chat (or leave it registered but unwired).

The messaging groups register / init-first-agent create default their unknown_sender_policy to whatever the channel adapter declares for that context (DM vs group) — strict when the channel has no declaration — so unknown senders are gated until you add them (or an admin approves the adapter-declared approval card). Pass --unknown-sender-policy to register to override. Show the user the OpenClaw allowlist and confirm who to grant before running the commands.

Phase 3: Identity and Memory

Fully conversational — read files directly and discuss. Placement depends on the Phase 1 choice:

• Shared identity: merge the same core identity/personality into every selected group's instructions.prepend.md.
• Fully separate / primary only: merge identity/personality only into the corresponding group's instructions.prepend.md.

Never edit a composed CLAUDE.md or AGENTS.md; it is regenerated each spawn. Put standing behavior in instructions.prepend.md and facts in memory/.

Find workspace files at <STATE_DIR>/workspace/. If AGENT_COUNT > 1, also check <STATE_DIR>/agents/*/workspace/ and ask which agent maps to which v2 agent group.

IDENTITY.md / SOUL.md

Read them. Distinguish always-loaded vs reference:

• Standing behavior (core traits, communication style, key rules) → weave into the group's instructions.prepend.md.
• Reference (backstory, extended guidelines) → a separate durable concept in an appropriate folder under groups/<folder>/memory/, linked from that folder's index.md and the root Map.

Choose each memory folder based on which related information will be easiest to find together; a folder may contain different concept types. Before writing the first concept into a new folder, create the folder and its index.md. Follow memory/system/definition.md, including its YAML frontmatter rules, for every new concept.

Show proposed edits before applying — this is a thoughtful merge, not a paste.

USER.md

Create a focused user-context concept in an appropriate memory folder and link it through that folder's index and the root Map. Put only facts relevant in nearly every conversation (for example name or timezone) into ## Core Memory; keep all other details in the linked file.

MEMORY.md and daily memory files

Show MEMORY.md; keep relevant items in focused concepts under the chosen memory folders, with links through each folder index and the root Map. For daily files (workspace/memory/*.md, count = DAILY_MEMORY_FILES):

AskUserQuestion: "You have N daily memory files. How to handle them?"

1. Copy as-is — agree on a descriptive folder, create it and its index.md, then copy with cp <workspace>/memory/*.md <group_dir>/memory/<chosen-folder>/ and link the retained files through its index and the root Map.
2. Consolidate — read, extract durable facts, and place them in focused linked memory files.
3. Skip.