horka-memory-restore

Restore Project Context

Scope

Reads and creates memory files only. Does NOT:

• Modify project source code
• Make architectural decisions
• Delete existing memory files (only adds/updates)

Step 1: Find memory files

Use Glob to search these locations in priority order:

1. .claude/memory/**/*.md
2. docs-architecture/**/*.md
3. CLAUDE.md, .claude/CLAUDE.md
4. documentations/**/*.md, docs/**/*.md
5. README.md

If ANY files found -> Step 2. If NOTHING found -> Step 3.

Step 2: Restore and summarize

Read each file found. Priority order:

File	Why it matters
PROJECT_STATE.md	Current phase, blockers, recent completions
ARCHITECTURE.md	Tech stack, structure, key flows
DECISIONS.md	Conventions and rationale behind them
NEXT_STEPS.md	What to work on next
COMMANDS.md	Dev commands to avoid re-discovery
CLAUDE.md	Project-level instructions

Output this structure:

## Context Restored

**Project**: [name from files or directory]
**Stack**: [technologies identified]
**Phase**: [current state]

### Done
- [completed items]

### Remaining
- [priority tasks]

### Watch Out
- [known bugs, blockers, important decisions]

---
What do we work on next?

Step 3: Initialize memory

When no memory files exist, create them automatically -- do NOT ask for permission, just analyze and create.

1. Project name: Use Glob for package.json, Cargo.toml, setup.py, *.xcodeproj. Fallback: run basename $(pwd) via Bash.
2. Tech stack: Use Glob for config files (*.config.*, Makefile, Dockerfile, *.toml). Read the ones found.
3. Current state: Run git log --oneline -10 via Bash. Read README.md if it exists.
4. Structure: Run ls -la via Bash for top-level layout.

Create .claude/memory/ with these files:

• PROJECT_STATE.md -- Current phase, done/remaining, known issues
• ARCHITECTURE.md -- Tech stack, structure, key components
• DECISIONS.md -- Conventions and key choices
• NEXT_STEPS.md -- Prioritized roadmap
• COMMANDS.md -- Useful dev commands