By skvcool-rgb
Local-first memory + reality-sync for Claude Code. Auto-injects MEMORY.md, git state, and session history. Multi-machine sync, marker-fenced MEMORY.md auto-curation, opt-in test runner, cross-tool surfaces. Pure-stdlib, zero dependencies.
Seed kos-memory from existing project docs + Claude Code transcripts (one-shot, idempotent)
Mine high-value chunks and append/refresh a marker-fenced suggestions block in MEMORY.md
Export project memory to portable JSON — for backup, migration, or sharing
Import a kos-memory JSON export into this project's store
Inspect or change kos-memory mode (primary | backup)
Admin access level
Server config contains admin-level keywords
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Local-first memory + reality-sync for any LLM coding tool. Auto-injects MEMORY.md, git state, and session history at session start. Bootstraps from existing project docs and Claude Code transcripts. Multi-machine sync, marker-fenced MEMORY.md auto-curation, opt-in test-runner integration. HTTP + MCP + CLI surfaces for Claude Code, Claude Desktop, Cursor, Cline, Zed, Aider, Continue.dev, and shell scripts. Pure-stdlib, zero dependencies.
v5 was Claude-Code-only with bounded retrieval and no project bootstrap. v6 fills six gaps:
| Gap | v6 closes it via |
|---|---|
| Cold start (empty store on fresh install) | /memory-bootstrap seeds from README/CHANGELOG/etc. + replays existing ~/.claude/projects/<encoded>/*.jsonl transcripts |
| MEMORY.md is operator-only / never auto-updates | /memory-curate writes high-value chunks (user-asserted, decision-pattern matches, version bumps) into a marker-fenced section operator reviews |
| Tree summary too shallow | tree_depth + tree_max_entries config keys in .kos-memory/config.json |
| No live test signal | lib.test_runner auto-runs collect-only on every survey (safe); full run opt-in via KOS_MEMORY_RUN_TESTS=1 |
| Single-machine, no sync | `/memory-sync push |
| Claude-Code-only | Standalone MCP stdio server + HTTP API on 127.0.0.1 + universal CLI; integration recipes for 7 tools |
[kos-memory PRIMARY] Memory reconstruction (1,247 chunks, 52 sessions, last ingest: today)
## MEMORY.md anchors (operator-curated truth)
### project_memory_md (MEMORY.md, 14m ago, 256892 bytes)
# 🚨 RESUME-HERE POINTER
# 🌐 LANGUAGE MIGRATION ROADMAP
...
## Live project state (filesystem + git, surveyed now)
branch: main (clean)
head: abc1234 "v6.0.0 — close all gaps"
vs upstream: origin/main (0 ahead, 0 behind)
tags: v6.0.0, v5.0.0, v4.1.0, ...
last commits:
abc1234 v6.0.0 — close all gaps
bbcaa92 feat(v6 baseline): configurable tree depth/breadth
2e3c6a2 docs(README): polish pass
versions:
.claude-plugin/plugin.json: 6.0.0
lib/__init__.py: 6.0.0
tree: .claude-plugin/, lib/ (16 .py), hooks/ (4 .py),
mcp/ (5 .py), commands/ (10 .md), tests/ (18 .py)
test runner: unittest, collect=374
## Recent session catalog (auto-extracted)
- 2026-05-04 · abc1234 [v6, sync, bootstrap, http] → ...
- (more)
## Build-status reconciliation (chunks vs filesystem)
✓ confirmed (chunks + filesystem agree): lib/store.py, hooks/Stop.py, ...
## Drift
- (no drift; MEMORY.md aligns with chunks)
Authority order for any claim about project state:
1. Live project state (filesystem + git) — ground truth
2. MEMORY.md anchors — operator-curated truth
3. User-asserted chunks (/remember) — explicit pins
4. Auto-extracted chunks — high-recall, may be stale
When you ask "is X built", the UserPromptSubmit hook auto-runs:
[reality check] 'X' is BUILT — confirmed by 12 chunks, filesystem evidence,
and git (commit abc123 "X release").
evidence: chunks=claimed_built, filesystem=confirms, git=committed (confidence: high)
Requirements: Python 3.9+ and Claude Code (or any MCP-compatible / HTTP-capable tool). Nothing else.
git clone https://github.com/skvcool-rgb/KOS-MemoryV4.git
cd KOS-MemoryV4
python scripts/install.py # use python3 on Mac/Linux if python isn't on PATH
Restart Claude Code. In any project, type /memory-status to verify, or /memory-bootstrap if the project already has a README/CHANGELOG/prior CC transcripts you want to seed memory with in one shot.
The installer:
~/.claude/settings.json (atomic write, with backup)..kos-memory/, ingests a chunk, recalls it).docs/integrations/).Idempotent. Re-run any time.
Same per-project .kos-memory/chunks.db — every tool that knows where to look reads from and writes to it. Integration configs in docs/integrations/:
npx claudepluginhub skvcool-rgb/kos-memoryv4 --plugin kos-memoryComprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
Permanent coding companion for Claude Code — survives any update. MCP-based terminal pet with ASCII art, stats, reactions, and personality.
Develop, test, build, and deploy Godot 4.x games with Claude Code. Includes GdUnit4 testing, web/desktop exports, CI/CD pipelines, and deployment to Vercel/GitHub Pages/itch.io.
Comprehensive PR review agents specializing in comments, tests, error handling, type design, code quality, and code simplification
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review