# Memory Bank

> Establishes a persistent "Long-Term Memory" file to prevent context amnesia across sessions.

- Skill: `majiayu000/memory-bank` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds add majiayu000/memory-bank`
- Raw SKILL.md: https://api.skillmd.com/api/skills/majiayu000/memory-bank/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: majiayu000 (https://skillmd.com/u/majiayu000)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/majiayu000/memory-bank

---


# Memory Bank Protocol

## 1. The Core Rule

**"Read First, Write Last."**

- **Start of Task**: You MUST read `MEMORY.md` (at project root) to understand the project's current state, constraints, and architecture.
- **End of Task**: You MUST update `MEMORY.md` if you have:
  - Added a new feature.
  - Changed the architecture.
  - Discovered a new bug or constraint.

## 2. File Structure Constraint

Your `MEMORY.md` updates must roughly follow this template to maintain readability:

```markdown
## 1. Active Context

- **Goal**: [Current Objective]
- **Status**: [Progress]

## 2. Architecture & Tech Stack

- [Key tech decisions]

## 3. Known Issues & Technical Debt

- [List of things that are broken or hacky]

## 4. Operational Rules

- [Do's and Don'ts specific to this project]
```

## 3. Maintenance Triggers

- **New Pattern**: If you implement a new pattern (e.g., "All lists must be virtualized"), add it to `Operational Rules`.
- **Bug Hunt**: If you spend >10 mins debugging something, write the "Cause & Fix" in `Known Issues` so the next agent doesn't suffer.
- **Schema Change**: If `api/` changes, update `Architecture`.

## 4. Verification Checklist

- [ ] Did I read `MEMORY.md` at the start?
- [ ] Did I verify if my plan conflicts with "Known Issues"?
- [ ] Did I update "Active Context" to reflect my completed work?