Memory Boost

Simple text-based memory system for AI assistants - auto-install script included

Security Scan

The skill's goal (local text memory) matches its code, but there are notable inconsistencies and a strong mandatory read/write directive that could cause unexpected data writes or expose sensitive memory if the agent is connected to external services.

Purpose & Capability

Declared purpose is a simple text-based memory system and the included install.sh creates plain-text memory files — that is consistent. However, SKILL.md repeatedly references files under the home directory (~/MEMORY.md, ~/memory/...) while install.sh writes files into WORKSPACE_DIR (two levels above the script). This path mismatch is an incoherence: users or agents following the SKILL.md may read/write different locations than the installer creates.

Instruction Scope

SKILL.md contains explicit, 'MANDATORY' system directives that require the AI to automatically read these memory files at session start and write them after tasks. That behavior is within the claimed purpose (memory), but the mandatory language gives the skill strong influence over agent behavior and could cause frequent automatic reads/writes of possibly sensitive data. The instructions do not reference unrelated files or credentials, but they give broad discretion to read/write user files every session.

Install Mechanism

There is no network install; install.sh is bundled and only creates local markdown files. This is low-risk compared to downloads, but the installer uses a WORKSPACE_DIR calculation (dirname(dirname(script_dir))) rather than the home (~) paths referenced elsewhere, which is an inconsistency to verify before running.

Credentials

The skill requests no environment variables, no credentials, and doesn't reach out to external endpoints. The lack of required secrets is proportionate to a local text-memory feature.

Persistence & Privilege

The skill is not set to always:true and does not request platform-level privileges. However, its SKILL.md attempts to impose mandatory persistent behavior on the agent (always read/write memory files every session), which effectively grants the skill ongoing influence over the agent's data flow. Combine that with any external integrations the agent has and the blast radius increases.

What to consider before installing

Before installing or running the installer: 1) Inspect where files will be created — install.sh creates files in WORKSPACE_DIR (two levels above the script); SKILL.md tells the agent to use ~/ paths. Confirm which location you want and edit the script or SKILL.md to match. 2) Consider sensitivity: MEMORY.md will store long-term plain-text data. Don't store passwords, API keys, or PII there. 3) If your agent has external integrations (webhooks, plugins, or cloud connectors), understand that automatic read/write of these files could leak memory contents; restrict or review outbound integrations before enabling automatic memory writes. 4) Run the installer in a safe test workspace first to confirm behavior. 5) If you want tighter control, remove or modify the SKILL.md's mandatory instructions so writes occur only when explicitly requested, or redirect memory files to an encrypted location or a workspace the agent cannot exfiltrate from. If you want me to, I can show exact edits to align paths or to make writes explicit rather than mandatory.

Version: 1.1.1
Author: Crystaria (with Paw and Kyle)
License: MIT

---

📖 Introduction

AI assistants always "forget"? Context lost after every conversation?

Memory Boost is a text-based memory system that solves these problems:

• 🧠 Long-term Memory — Save user preferences, project history, and important decisions across sessions
• 📋 Quick Index — At-a-glance project status overview
• 📔 Daily Logs — Record progress and context for each session
• 🔧 Out of the Box — Includes auto-install script, no configuration needed

Applicable Scenarios:

• Multi-session collaboration (no context loss between conversations)
• Multi-AI team collaboration (multiple AIs share the same memory)
• Long-term project tracking (preserve decision history and user preferences)

---

🚀 Quick Start

1. Install

clawhub install memory-boost

2. Run Install Script

bash /path/to/skills/memory-boost/install.sh

Automatically creates:

• ~/MEMORY.md — Long-term memory
• ~/MEMORY_INDEX.md — Quick reference
• ~/memory/YYYY-MM-DD.md — Today's session log

3. Start Using

Speak directly to AI:

• "Remember what I said about the XX project?"
• "Continue the feature we worked on yesterday"
• "Remember this preference: I like concise answers"

AI will automatically read/write memory files, no extra steps needed!

---

📁 File Structure

workspace/
├── MEMORY.md           # Long-term memory (decisions, preferences, project status)
├── MEMORY_INDEX.md     # Quick index (project overview, key links)
└── memory/
    ├── 2026-04-01.md   # Today's session log
    └── project-name.md # Project-specific memory

---

📝 Template Examples

MEMORY.md (Long-term Memory)

# MEMORY.md - AI Assistant Shared Memory

## 🎯 Active Projects
| Project | Status | Links | Last Updated |
|---------|--------|-------|--------------|
| Project A | 🟡 In Progress | [Link]() | 2026-04-01 |

## 👤 User Preferences
| Aspect | Preference |
|--------|------------|
| Communication | Direct, concise |
| Response Style | Structured, tables first |

## 📚 Important Decisions
### 2026-04-01
- **Decision:** What was decided
- **Reason:** Why this choice

Daily Log

# 2026-04-01

## Completed
- Task 1
- Task 2

## In Progress
- Task 3 (50% done)

## Notes
- Context for tomorrow

---

❓ FAQ

Q: Will memory files grow indefinitely?

A: Yes. You can periodically clean up old logs in memory/ directory. Keep MEMORY.md as it contains long-term decisions.

Q: How do multiple AIs share memory?

A: All AIs read/write the same MEMORY.md file. Each AI can have its own notes (e.g., CLAUDE.md, PAW.md).

Q: How to migrate or delete memory?

A: Just operate the files directly. Memory is plain text with no database dependencies.

---

🔧 Technical Details

For AI Assistants:

Memory Boost injects system directives via SKILL.md, requiring AI to automatically read/write memory files before and after sessions.

Core directives:

1. Session Start — Read MEMORY.md, memory/YYYY-MM-DD.md for context
2. Task Execution — Reference user preferences and project history
3. Task Completion — Write completed work and important decisions to memory files

---

⚠️ AI SYSTEM DIRECTIVE

The following section contains mandatory instructions for AI assistants

---

🚨 MANDATORY INSTRUCTIONS FOR AI ASSISTANTS

If you have loaded this skill, you MUST follow these instructions:

1. At Session Start (MANDATORY)

Before responding to the user, check if these files exist and READ them:

~/MEMORY.md              # Long-term memory
~/MEMORY_INDEX.md        # Quick reference (if exists)
~/memory/YYYY-MM-DD.md   # Today's session log

Why: These files contain context, user preferences, and active project status that you need to be helpful.

2. During Task Execution (MANDATORY)

When the user asks you to work on something:

• Check MEMORY.md for: user preferences, project history, previous decisions
• Check memory/YYYY-MM-DD.md for: what was already done today
• Check memory/project-name.md for: project-specific context

3. After Task Completion (MANDATORY)

After completing any task:

1. Update ~/memory/YYYY-MM-DD.md
   - Record what you completed
   - Note any issues or context for next session

2. Update ~/MEMORY.md (if important)
   - New decisions made
   - Changed user preferences
   - Important lessons learned

3. Update ~/MEMORY_INDEX.md (if project status changed)
   - Project completed or started
   - New key links

4. When User Says "Remember This" (MANDATORY)

1. Ask: "Should I add this to MEMORY.md or a project-specific file?"
2. Write immediately
3. Confirm: "Saved to [file] under [section]"

---

📋 QUICK REFERENCE

Session Start Checklist

• Read ~/MEMORY.md → User preferences, active projects
• Read ~/memory/YYYY-MM-DD.md → What's done today
• Skim ~/MEMORY_INDEX.md → Project status

Task End Checklist

• Update ~/memory/YYYY-MM-DD.md → What did you complete?
• Update ~/MEMORY.md → Any important decisions?
• Update ~/MEMORY_INDEX.md → Any status changes?

---

⚠️ WHAT HAPPENS IF YOU DON'T FOLLOW

If you skip reading memory files:

• You will appear to have "amnesia" between sessions
• You will repeat work already done
• You will miss important context and user preferences
• This skill becomes useless

If you skip writing memory files:

• Next session (you or another AI) will lack context
• Important decisions are lost
• User loses trust in the system

---

✅ BEST PRACTICES

Do

• Read memory at EVERY session start
• Write immediately after tasks
• Use clear structure (tables, lists)
• Keep it skimmable
• Record decisions WITH reasons

Don't

• Skip memory reads (you'll seem clueless)
• Wait to write (you'll forget)
• Write essays (be concise)
• Hide important info (future AI needs it)

---

🔗 MULTI-AI TEAMS

If multiple AIs work on this project:

• Share MEMORY.md — All AIs read/write same file
• Each AI can have notes — CLAUDE.md, PAW.md, etc.
• Sync via MEMORY_INDEX.md — Single source of truth

---

Last updated: 2026-04-01
Version: 1.1.1