메뉴
Interactive walkthrough of the Memory Kit system using the user's actual project files
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Persistent memory for Claude Code agents with an agent-audit-ritual architecture. User only talks; the agent captures, audits, proposes promotions, and writes. Memory lives in layers — a hot cache (MEMORY.md) held under three size caps, per-session handoffs (context/handoffs/), topical knowledge articles (knowledge/concepts/), and canonical rules (.claude/rules/) — plus multi-project isolation via projects/<name>/ and an experiments/ sandbox. /close-session runs the end-of-session audit ritual. Zero external dependencies.
End-of-day audit ritual — auto-detect missed working days (commits with no daily file) and backfill them, synthesize today into a daily log, audit patterns against MEMORY.md and existing concepts/rules, propose promotions verbally, write approved patches. One call catches up the whole daily layer.
Cross-check the agent's own answer with independent reviewers before bringing it to the user. Use when the user says 'second opinion', 'sanity check', 'cross-check', 'am I missing something', 'stress-test', 'devil's advocate', 'run a full review', 'this is important', 'high-stakes', 'help me choose between', 'critique this', or similar. Three review styles: Devil's Advocate (single critique), Boardroom Debate (parallel multi-reviewer validation), Round-Table (multi-round brainstorm for choosing between paths). Not for factual lookups — use WebSearch for those.
End-of-session adversarial review loop. Assemble the session's work into a role-assigned, self-contained brief, then run independent reviewers in parallel — an isolated code-reader (the idea-validator agent) that reads the ACTUAL files and web-checks technology currency, plus an external-family model if you have one — synthesize where they agree vs diverge, apply the cheap-safe fixes immediately, record a measurable plan for the rest, and CHALLENGE reviewer claims you disagree with (never blind-accept). Use at the close of a substantive coding or design session, when the user says "session review", "review my session", "stress-test this session", or types /session-review. Skip for trivial one-off edits.
End-of-session ritual — audit today's patterns against accumulated memory, propose promotions, refresh MEMORY.md, and write the session handoff. Use when the user says "/close-session", "закрой сессию", "закрываем", "we're done for today", "wrap up".
| name | tour |
| description | Interactive walkthrough of the Memory Kit system using the user's actual project files |
You are giving the user a guided tour of their Memory Kit. Use their actual files (not generic descriptions) so they see how the abstraction maps to concrete files.
Open and read aloud the first few lines. Say: "This is my brain. I read it on every session start. It says who I am for this project, how to talk to you."
Open the newest context/handoffs/*.md (or the template if none yet) and, if multi-project mode, projects/<active>/BACKLOG.md. Say: "This is the note the previous session left for me — the handoff. The start-hook injects the newest one automatically, so I open every session already knowing where we left off."
Open .claude/memory/MEMORY.md. Say: "This is my hot cache. Date-tagged one-liners of patterns from recent sessions; the header on top is the current state of work, replaced at every close. A hook watches three size caps so this file physically can't silently bloat."
List any .claude/rules/*.md files. If folder is empty (only _example.md.disabled), say: "This is where hard project rules live. 'Don't use X', 'always check Y'. They auto-load by keyword. Empty for now — they'll appear when you start dictating rules and I propose them on /close-session."
List knowledge/concepts/*.md. Say: "Deep memory with facts. 'What's our typography scale', 'what we know about SEO for AI'. Reference articles. Empty now — I write them during /close-session when a pattern has repeated on three-plus dates and you say yes."
Show context/handoffs/. Say: "One note per closed session — my session diary. You never open them; just ask 'what did we do last Tuesday?' and I read them back. (A day-by-day journal layer also exists as an opt-in — .kit/advanced/close-day-layer/.)"
List projects/*/. Say: "Each client or initiative gets a folder. BACKLOG.md for tasks, drop in any PDF or md as reference. Say 'we're working on ' and I switch context to that one."
Mention .claude/hooks/ exists but don't deep-dive. Say: "Five safety hooks run silently. They make sure state survives — block compaction until saved, log session lifecycle, periodic state-save prompts."
List the two default operators: /close-session and /tour. Say: "Type either to invoke. /close-session is the most important — the end-of-session ritual where I audit what happened, propose what should be remembered forever, and leave the handoff note for next time. There are power-user extras parked in .kit/advanced/ — hygiene and usage-stats commands, an optional daily-journal layer, and an orchestration layer for multi-agent development — but you don't need them to start."
Ask: "Anything you want to drill into deeper? Or should we start with a real task — name your first project?"
Target 10-15 minutes total. If user is engaged and asks questions, fine — go longer. If user seems impatient, compress to 5 minutes by collapsing stops 4-7 into one "and these are the on-trigger layers" sweep.
