菜单
4-tier memory system management for all stateful agents — loading, saving, pruning, promoting, and maintaining memory files. Use when user says "load memory", "save memory", "update working memory", "prune memory", "memory management", "tier system", or any agent memory lifecycle activity.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/htekdev/copilot-home-assistant --skill memory-management命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | memory-management |
| description | 4-tier memory system management for all stateful agents — loading, saving, pruning, promoting, and maintaining memory files. Use when user says "load memory", "save memory", "update working memory", "prune memory", "memory management", "tier system", or any agent memory lifecycle activity. |
Standard 4-tier memory system used by all stateful agents in the {{FAMILY_NAME}} family platform. This skill defines the canonical patterns for loading, saving, pruning, and promoting agent memory.
| Tier | File | Purpose | Load Rule | Size Limit |
|---|---|---|---|---|
| 1 | core.md | Identity, mission, heuristics | ALWAYS load first | 3-5KB |
| 2 | working.md | Current state, active context | ALWAYS load second | 5KB max |
| 3 | long-term.md | Historical patterns, lessons | On-demand only | 10KB max |
| 4 | events.log | Append-only event stream | Never bulk-load | Unlimited (prune >30 days) |
All agent memory lives at:
data/agents/{agent-name}/
├── core.md # Tier 1 — identity, rules, preferences
├── working.md # Tier 2 — current state, today's context
├── long-term.md # Tier 3 — accumulated wisdom, patterns
└── events.log # Tier 4 — append-only audit trail
MANDATORY for every stateful agent at run start:
1. Read data/agents/{agent-name}/core.md # Tier 1 — ALWAYS
2. Read data/agents/{agent-name}/working.md # Tier 2 — ALWAYS
Do NOT bulk-load Tier 3 (long-term.md). Only read it when:
Never load Tier 4 (events.log) in bulk. It's write-only during normal operations.
MANDATORY for every stateful agent before ending a run:
Update data/agents/{agent-name}/working.md with:
Format the timestamp:
## Last Updated
2026-05-06T07:30:00-05:00
Size discipline: Keep working.md under 5KB. When approaching the limit:
Append one-line entries to data/agents/{agent-name}/events.log:
[2026-05-06T07:30:00-05:00] action: description of what happened
Rules:
scan:, create:, fix:, notify:, skip:Append to data/agents/{agent-name}/long-term.md ONLY when:
Do NOT promote:
A working memory file is STALE if:
Last Updated timestamp is >3 days old AND the agent has an active cron jobWhen stale memory is detected:
# {Agent Name} — Core Identity
## Last Updated
{ISO timestamp}
## Identity
{1-2 sentences: who this agent is}
## Mission
{2-3 bullet points: what it exists to do}
## Ownership Boundaries
### You own
- {list of owned domains}
### You do NOT own
- {explicit exclusions}
## Core Heuristics
1. {Key decision rule 1}
2. {Key decision rule 2}
...
## Key Rules
- {Critical behavioral rules}
# {Agent Name} — Working Memory
## Last Updated
{ISO timestamp}
## Current State
{What's active right now — 3-5 bullet points max}
## Recent Actions
{What happened in last 1-3 runs}
## Pending / Deferred
{Items waiting for input or scheduled for later}
## Active Rules
{Any temporary rules or sprint-mode adjustments}
All agents with persistent state use this memory system:
Stateless agents (no memory): daily-briefing, meal-planner, weekly-planner, budget-review, heartbeat, checkin, family-coordinator