atsarevskiy

Claude Knowledge Sync — Development skill for Claude Code

Development community

An environment-aware Claude Code skill that synchronizes and merges memories and steering documents across multiple machines, preserving each machine’s unique tools, workflows, and local context.

How to install Claude Knowledge Sync

This entry records only its repository, not the path inside it, so there is no exact command to give. Open atsarevskiy/claude-knowledge-sync and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Claude Knowledge Sync does

An environment-aware Claude Code skill that synchronizes and merges memories and steering documents across multiple machines, preserving each machine’s unique tools, workflows, and local context.

Alternatives in Development

  • Code Simplifier — Simplify code while preserving functionality 14k ★
  • Dotenv Vault — Sync environment variables across machines for consistent Claude Code behavior 1.2k ★
  • Caveman Claude Skill — Ultra-compressed communication mode for Claude 396 ★

README

Claude Knowledge Sync

An environment-aware Claude Code skill that synchronizes and non-destructively merges knowledge across machines. It uses Claude Code's actual memory and instruction files, shares explicitly portable facts, preserves conflicts, and keeps snapshots, audit logs, and classification history.

Install from the Claude Code marketplace

claude plugin marketplace add atsarevskiy/claude-knowledge-sync
claude plugin install knowledge-sync@claude-knowledge-sync

Native Claude Code documents

Knowledge Sync does not create a competing memory or steering document. It discovers existing Markdown files in:

  • ~/.claude/CLAUDE.md and ~/.claude/rules/**/*.md for user steering;
  • ~/.claude/memory/**/*.md, if a user maintains that location;
  • ~/.claude/projects//memory/*.md fact files. MEMORY.md is a generated local index and is deliberately excluded from sync.

Only `~/.claude/knowledge-sync/` is created, for this plugin's metadata. `init` is valid even when no native document exists. When a document is later discovered, it is added to the configuration; when one disappears, it is reported as missing rather than recreated or removed.

Portable and local facts

Per-fact memory files use Claude's frontmatter scope. A whole fact file is portable when `metadata.scope` is `common`; any other label keeps it local:

---
metadata:
  scope: common
---

# Test workflow
Run focused tests before the full suite.

For free-form steering documents such as `CLAUDE.md` and rule files, use inline `` blocks. Mining reads and records the fact-file scope plus source hash. It never moves, replaces, or erases existing memory.

Bootstrap and sync

knowledge_sync.py init --machine A
knowledge_sync.py add-peer --name B --ssh-target you@host
# Or provision the remote helper and add the peer in one direction:
knowledge_sync.py pair --ssh-target you@host

`pair` copies only the helper and creates its `~/.claude