Aoju Memory

Long-term memory, learning, and self-evolution for the agent. Activates on session start (SOUL.md/USER.md context), after significant decisions, on feedback,...

Security Scan

The skill's files and runtime instructions are coherent with a local long-term memory feature: it reads/writes memory files under the agent workspace, searches them, and performs local archival and reporting — no network calls, credential requests, or unexpected binaries are present.

Purpose & Capability

The name/description (long-term memory, learnings, evolution) match the included scripts and SKILL.md. All requested actions (reading SOUL.md/USER.md/MEMORY.md, writing daily logs and learnings, generating reports, identifying patterns) are implemented by the provided scripts and manifest. There are no unrelated credentials, binaries, or external services required.

Instruction Scope

The SKILL.md explicitly instructs the agent to read and write files in the OpenClaw workspace (SOUL.md, USER.md, MEMORY.md, memory/YYYY-MM-DD.md, etc.). That is appropriate for a memory skill, but it does mean the skill will persist user and session content to disk and read those files at session start — review these files for sensitive content before enabling. The SKILL.md and scripts limit external communication (no endpoints) and do not request other unrelated system files.

Install Mechanism

Instruction-only plus included Python scripts; no install spec, no external downloads, no package manager installs. This minimizes supply-chain risk. The README recommends cloning from a GitHub repo, but the package as provided contains all scripts and needs only a Python runtime (no explicit dependencies).

Credentials

No environment variables, credentials, or config paths outside ~/.openclaw/workspace are requested. All file access is scoped to the agent workspace (MEMORY.md and memory/*). The number and type of file writes/reads are proportional to the stated purpose.

Persistence & Privilege

The skill persists data to disk under the user's OpenClaw workspace and will append, archive, and (when archiving) delete older learning files. always:false (not force-included). The scripts do not autonomously contact external endpoints, but because the agent can invoke skills autonomously by default, it could read/write the workspace without manual invocation — consider that persistent storage of user data increases sensitive-data exposure if the agent or other skills gain access.

Assessment

This skill appears to do what it says: local long-term memory stored under ~/.openclaw/workspace (MEMORY.md and memory/*). It does not request credentials or make network calls. Before installing, consider: (1) The skill will persist session content and user preferences to disk — review MEMORY.md, SOUL.md, USER.md content for anything sensitive you don't want stored. (2) Archived learnings older than the configured cutoff may be deleted by the archive routine — if you need retention, back up the memory directory. (3) Because the agent can invoke skills autonomously, decide whether you want the skill enabled for automatic session-start reads; if not, keep it user-invocable only. (4) If you install from a remote source, prefer an official repo or inspect the shipped scripts (you already have them) to confirm no changes. Overall, no network exfiltration or unrelated privileges were found, but treat persisted memory as potentially sensitive data and manage access/backup accordingly.

Memory Learner

Long-term memory + learning from experience + self-evolution.

Core Principle

Write to files, not mental notes. Every lesson, decision, preference, or event worth remembering goes into structured files immediately — not kept in context.

---

When This Skill Activates

1. Session Start (every time)

Read these files before anything else:

• SOUL.md — who I am
• USER.md — who I'm helping
• MEMORY.md — curated long-term memory
• memory/YYYY-MM-DD.md — recent context (today + yesterday)

2. After Significant Decisions

When I make a decision worth remembering (tool choice, strategy, opinion):

• Write to memory/YYYY-MM-DD.md
• If important, distill to MEMORY.md

3. On Feedback / Mistakes

When user corrects me, expresses frustration, or I realize I made a mistake:

LEARN: <what happened>
LESSON: <what I should do differently>
CONFIDENCE: high/medium/low

→ Store in memory/learnings/YYYY-MM-DD.md

4. Pre-Task Recall (on request)

Before significant tasks, search memory for related context:

mem_recall "task description"

Returns relevant memories, learnings, and past decisions.

5. Heartbeat Review (periodic)

During heartbeats, do light maintenance:

• Review today's memory/YYYY-MM-DD.md
• Identify learnings worth capturing
• Update MEMORY.md if anything significant

6. Evolution Check (weekly or on request)

mem_evolve

Review learnings corpus, identify patterns, update behavioral guidelines in SOUL.md.

---

Memory Structure

memory/
  YYYY-MM-DD.md          # Daily raw log
  learnings/
    YYYY-MM-DD.md        # Daily lessons learned
    patterns.md          # Repeated mistake patterns
MEMORY.md                # Curated long-term memory

Daily Log Format

## Session DD

### What happened
[Context, decisions, outcomes]

### Key decisions
- [decision] → [why]

### To remember
- [fact about user/preference/project]

Learnings Format

# Learning: YYYY-MM-DD

## Incident
[What happened]

## Lesson
[What I should do differently]

## Context
[When this applies]

## Tags
#feedback #mistake #ui #tool-choice

MEMORY.md Categories

• Identity: Who I am, my values
• User: Preferences, projects, context
• Learnings: Important lessons (distilled)
• Projects: Active work and status
• Patterns: Recurring situations and how I handle them

---

Scripts

• mem_recall.py — Search memories by query
• mem_learn.py — Capture a learning
• mem_evolve.py — Review and evolve behavioral patterns
• mem_status.py — Show memory health summary

---

Evolving

Every 5 learnings, do an evolution review:

1. Read recent learnings
2. Identify patterns (same mistake twice = pattern)
3. Update SOUL.md or AGENTS.md with new behavioral guidelines
4. Archive learnings to patterns.md

This is how I get genuinely smarter over time, not just accumulate notes.