migrate-memory

Migrate agent memory between provider stores after a group provider switch.

2|Updated Jan 14, 2026
One-click install
npx skills add https://github.com/RADHA0-max/selfhealbackend --skill migrate-memory-radha0-max
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: migrate-memory
Source: https://github.com/RADHA0-max/selfhealbackend/tree/main/nanoclaw/.claude/skills/migrate-memory
Command: npx skills add https://github.com/RADHA0-max/selfhealbackend --skill migrate-memory-radha0-max

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When an operator switches a NanoClaw agent group from one provider to another (e.g. Claude to Codex), the agent loses all memory because each provider keeps its own store. This Skill carries durable memory across the switch so the agent retains its identity, instructions, and learned facts. ## Core Features & Use Cases - Bidirectional migration: Converts memory from a flat CLAUDE.local.md file into a memory/ scaffold tree, or distills a scaffold tree back into CLAUDE.local.md, depending on the direction of the provider switch. - Distillation, not dumping: Infers what is durable (identity, seed instructions, user preferences, durable facts) and drops conversational residue, keeping the target store clean. - Safe and idempotent: Copies rather than moves (the source store remains as rollback), detects prior imports to avoid duplication, and restarts the group with ncl groups restart for verification. - Use Case: After running ncl groups config update --provider codex, ask the agent to migrate memory so the Codex-based group still knows the user's preferences and project context previously stored in CLAUDE.local.md. ## Quick Start Ask the agent to migrate the group's memory from the old provider to the new one after switching providers with ncl groups config update.

Frequently Asked Questions about migrate-memory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I migrate agent memory after switching providers in NanoClaw?

Run this Skill after switching the provider with ncl groups config update --provider. The agent reads the source provider's memory store, distills durable content into the target provider's store, and restarts the group with ncl groups restart.

How to carry memory from Claude to Codex in an agent group?

The Skill converts the flat CLAUDE.local.md file into the memory/ scaffold tree, writing distilled content to memory/memories/imported-agent-memory.md and adding an index entry. The reverse direction integrates memory/ content into CLAUDE.local.md under a marked section.

Does switching between two scaffold-based providers require memory migration?

No. Two providers that both use the memory/ scaffold share the same tree, so nothing needs to be migrated. Migration is only required when the switch crosses store shapes: flat to scaffold or scaffold to flat.

What happens to the old memory store after migration?

The source store is left exactly as it is, since the Skill copies rather than moves content. The untouched source acts as the rollback if anything goes wrong with the migration.

Why did my agent forget everything after a provider switch?

NanoClaw does not migrate memory at runtime; each provider keeps its own store. If the switch crossed store shapes and no migration ran, the new provider starts empty. Run the migration and verify with a test message referencing a migrated fact.