功能描述

为 OpenClaw 打造的五层自适应记忆检索系统，四层自进化记忆模型（L1轨迹→L2模式→L3世界模型→技能结晶），反馈驱动的价值评分，自动维护与健康诊断。适用场景：(1) 带降级保障的记忆搜索，(2) 带历史故障诊断的健康检查，(3) 每日散会流程自动模式归纳，(4) 生成启动简报替代全量MEMORY.md注...

使用说明 (SKILL.md)

MemCore 记忆核心

Name: MemCore 记忆核心
Author: adchina2025

为 OpenClaw 打造的五层自适应记忆系统，自进化。替代全量 MEMORY.md 注入 — 节省 92% token。

Five-tier adaptive memory system with self-evolution for OpenClaw. Drop-in replacement for full MEMORY.md injection — 92% token savings.

Quick Commands

# Startup: generate brief (≤500 tokens, replaces full MEMORY.md)
python3 scripts/memcore/cli.py brief

# Search: 5-tier adaptive fallback
python3 scripts/memcore/cli.py search "\x3Cquery>" -n 5

# Daily maintenance (index new logs → induce patterns → decay → refresh brief)
python3 scripts/memcore/cli.py run-all

# Stats overview
python3 scripts/memcore/cli.py stats

# Feedback: log whether retrieved memory was useful
python3 scripts/memcore/cli.py feedback-log \x3Ctrace_id> used|skipped|good|bad

Startup Integration

On each session start, replace full MEMORY.md loading with:

1. Read SOUL.md → USER.md → MEMORY_BRIEF.md (auto-generated) → .anatomy.md → yesterday+today memory
2. MEMORY.md loaded on-demand only via: python3 scripts/memcore/cli.py search "\x3Ckeywords>"

The brief generator picks top 5 active patterns, top 3 recent lessons, and current taskboard items — all under 500 tokens.

散会 Integration

Add step 3 to 散会 flow:

python3 scripts/memcore/cli.py index && \
python3 scripts/memcore/cli.py induce && \
python3 scripts/memcore/cli.py feedback

This auto-indexes today's log, induces new patterns, and decays stale traces.

Health Check Integration

When running daily health checks (09:00), append MemCore diagnostic:

# Get system stats
python3 scripts/memcore/cli.py stats

# If anomalies found, search historical similar faults
python3 scripts/memcore/cli.py search "\x3Canomaly keywords>" -n 3

Safety rule: Report only. Never auto-fix. Let human decide.

Search & Feedback Flow

After every memory_search call, log feedback:

# If the retrieved memory was helpful:
python3 scripts/memcore/cli.py feedback-log \x3Ctrace_id> used

# If irrelevant:
python3 scripts/memcore/cli.py feedback-log \x3Ctrace_id> skipped

This trains the retrieval system — high-value traces rise, stale ones decay.

Cron Setup

Two 12-hour maintenance jobs (systemEvent, main session, wakeMode: next-heartbeat):

Time (CST)	Cron expression	Action
04:00	`0 4 * * *`	index + induce + feedback + brief
16:00	`0 16 * * *`	same

These are SQLite-only, no model calls, near-zero resource cost.

Rollback

cp memcore_backup_\x3Cdate>/AGENTS.md ~/.openclaw/workspace/
cp memcore_backup_\x3Cdate>/MEMORY.md ~/.openclaw/workspace/

Original MEMORY.md is never modified. All MemCore data lives in separate SQLite files under ~/.openclaw/.

Detailed Reference

Architecture: See references/architecture.md for L1-L4 model, 5-tier retrieval details, file structure, and CLI reference.
EXO1 Upgrade: See references/upgrade-guide.md for Syncthing-based deployment to EXO1.

安全使用建议

Install only if you are comfortable with MemCore indexing your OpenClaw memory/workspace files into local SQLite databases and generating a startup MEMORY_BRIEF.md. Review the generated brief and any scheduled cron/systemEvent jobs, and avoid storing secrets or highly sensitive information in files that MemCore indexes.

功能分析

Type: OpenClaw Skill Name: memcore-memory Version: 1.0.3 MemCore is a legitimate memory management and optimization system for OpenClaw agents, designed to reduce token usage by replacing full memory logs with a structured retrieval system. The bundle uses SQLite databases to index agent traces and patterns, and includes a multi-tier search engine in `retriever.py` that safely utilizes `subprocess.run` for local file searching. No evidence of data exfiltration, malicious persistence, or harmful prompt injection was found; all operations are local to the agent's workspace and follow the stated purpose of enhancing memory efficiency.

能力评估

ℹ Purpose & Capability

The file reads OpenClaw memory/workspace materials, indexes them into SQLite, and generates a startup brief; this is aligned with the stated memory-retrieval purpose, but it handles potentially private agent memory.

ℹ Instruction Scope

The instructions ask users to integrate MemCore into startup, search feedback, daily wrap-up, and health-check workflows. This is disclosed and purpose-aligned, but it means retrieved or summarized memory can influence future sessions.

ℹ Install Mechanism

Registry metadata says there is no install spec, but the package includes a user-run install script that copies Python modules and, if memory logs exist, runs initial indexing and brief generation. This is not hidden, but users should run it knowingly.

ℹ Credentials

The shown code stays local under ~/.openclaw/workspace and ~/.openclaw SQLite files, with no credentials or network endpoints evident in the provided artifacts; the access is still broad within the OpenClaw memory/workspace context.

ℹ Persistence & Privilege

The skill documents optional cron/systemEvent maintenance and persistent SQLite databases. This is appropriate for a memory system, but users should decide whether scheduled background updates are acceptable.

版本历史

v1.0.3

memcore-memory 1.0.3 - No file changes detected in this version. - No updates to documentation, code, or functionality. - Version bump only; behavior remains unchanged.

v1.0.2

- Metadata updated: displayName field removed from SKILL.md. - No functional or instructional changes; usage and documentation remain the same.

v1.0.1

No user-facing changes in this version. - No file changes were detected. - Functionality and documentation remain unchanged.

v1.0.0

- Initial release of MemCore: a five-tier adaptive memory retrieval system for OpenClaw agents. - Replaces full MEMORY.md injection with automated startup briefs, saving up to 92% tokens. - Features self-evolving four-layer memory model, feedback-driven scoring, auto-maintenance, and health diagnostics. - Provides CLI tools for searching memory with fallback, daily pattern induction, health checks, and feedback logging. - Integrates with startup, 散会, and health check workflows; easy rollback and minimal resource cost.

元数据

Slug memcore-memory

版本 1.0.3

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 4

常见问题