remember

/remember - Memory Bank System

$ARGUMENTS

---

Purpose

This command creates or updates the project's Memory Bank — a set of persistent context files that survive across sessions. Claude reads these at session start to avoid re-discovering the same information.

---

Sub-commands

/remember              - Full scan: create/update all memory files
/remember context      - Update only active context (what's being worked on now)
/remember patterns     - Update only detected patterns and conventions
/remember decisions    - Update only architecture decisions
/remember issues       - Update only troubleshooting log
/remember status       - Show current memory bank status

---

Behavior

When /remember is triggered:

1. Scan Project

Auto-collect live context first using tools:

1. Branch: Run git branch --show-current
2. Recent commits: Run git log --oneline -10
3. Changed files: Run git diff --name-only HEAD~5

Then analyze the project to understand:

• Tech stack: package.json, requirements.txt, go.mod, Cargo.toml, etc.
• Architecture: folder structure, naming conventions, file organization
• Patterns: recurring code patterns, import styles, state management approach
• Recent activity: git log, recent changes, current branch

2. Create/Update Memory Files

Write findings to .agent/.claude/memory/:

File	Content
MEMORY-activeContext.md	Current work: branch, recent commits, active feature, blockers
MEMORY-patterns.md	Numbered patterns: naming conventions, file structure, code style, component patterns
MEMORY-decisions.md	Architecture Decision Records (ADR): each decision with rationale and date
MEMORY-troubleshooting.md	Solved issues: problem, root cause, solution, date

3. Preserve History

CRITICAL RULES:

• NEVER delete existing entries — only append or update
• Preserve all historical records (ADRs, past patterns, solved issues)
• When updating, add new entries with current date
• Mark outdated patterns as [DEPRECATED] instead of removing