심화 분석

아키텍처 분석, 내부 동작, 에코시스템 동향

7 articles

Claude Code Onboarding From Task Loops to MCP Policy

The repo is worth saving because it gives teams an onboarding system from the first task loop through Skills, subagents, hooks, and MCP.

2026년 6월 1일

claude codecodexonboarding

심화 분석11 min read

Claude Code From Scratch as an Agent Runtime Map

claude-code-from-scratch is worth saving as an agent runtime map: skills, compaction, subagents, worktrees, permissions, and MCP each get a concrete source-backed checkpoint.

2026년 6월 1일

claude codeagent architectureskills

심화 분석10 min read

iai-mcp for Local Agent Memory Boundaries

The useful question is not whether an assistant can remember more. It is whether a developer should let an MCP server capture every turn, store it locally, and inject recall into future sessions.

2026년 6월 1일

mcpmemoryclaude code

심화 분석10 min read

Make Claude Team Memory Reviewable

claude-teams-brain combines Claude Code hooks, SQLite/FTS5, MCP command-output filters, and reviewable memory commands. The useful question is not whether agents remember more, but whether the team can trust what gets injected.

2026년 6월 1일

claude codeagent memorymcp

심화 분석10 min read

Audit MCP Tool Risk with AgentSeal

AgentSeal can give Claude Code and Codex users a source-backed way to inspect MCP configs, poisoned tool descriptions, skill files, prompt probes, registry signals, and local agent paths before adopting a new tool.

2026년 6월 1일

agentsealmcpclaude code

심화 분석8 min read

Audit Agent Vibes Before Routing Cursor or Claude Code

Agent Vibes deserves a magazine rescue, but not as vendor-lock-in hype. The useful article is a cautionary architecture guide for routing Cursor, Claude Code, Codex credentials, and alternative backends through a local protocol bridge.

2026년 6월 1일

agent vibesclaude codecodex

vibe4g 위클리 구독

엄선한 바이브코딩 인사이트를 매주 목요일에 보내드립니다.

스팸 없음. 언제든 구독 해지.

심화 분석7 min read

Audit Claude Code MCP Calls with mcp-from-scratch

mcp-from-scratch is not a faster way to ship an MCP server. Its value is that it exposes the lifecycle, stdio, tools/list, and tools/call boundaries that Claude Code users need to review before trusting a tool.

2026년 6월 1일

mcpclaude codeprotocol