Memory
by @ivangdavila
Infinite organized memory that complements your agent's built-in memory with unlimited categorized storage.
clawhub install memoryπ About This Skill
name: Memory slug: memory version: 1.0.2 homepage: https://clawic.com/skills/memory description: Infinite organized memory that complements your agent's built-in memory with unlimited categorized storage. changelog: Redesigned as complementary system, user-defined categories, optional sync from built-in memory. metadata: {"clawdbot":{"emoji":"π§ ","requires":{"bins":[]},"os":["linux","darwin","win32"]}}
Memory π§
Superpowered memory that never forgets.
Your agent has basic built-in memory. This skill adds infinite, perfectly organized memory for everything else β parallel and complementary, never conflicting.
How It Works
Built-in Agent Memory This Skill (~/memory/)
βββββββββββββββββββββββ βββββββββββββββββββββββββββββββ
β MEMORY.md β β Infinite categorized storage β
β memory/ (daily logs)β + β Any structure you want β
β Basic recall β β Perfect organization β
βββββββββββββββββββββββ βββββββββββββββββββββββββββββββ
β β
Agent basics Everything else
(works automatically) (scales infinitely)
Not a replacement. Your agent's built-in memory keeps working. This adds a parallel system for unlimited, organized storage.
Setup
On first use, read setup.md to configure the memory system with the user. Key decisions:
1. What categories do they need?
2. Should we sync anything from built-in memory?
3. How do they want to find things?
When to Use
User needs organized long-term storage beyond basic agent memory: detailed project histories, extensive contact networks, decision logs, domain knowledge, collections, or any structured data that grows over time.
Architecture
Memory lives in ~/memory/ β a dedicated folder separate from built-in agent memory.
~/memory/
βββ config.md # System configuration
βββ INDEX.md # What's stored, where to find it
β
βββ [user-defined]/ # Categories the user needs
β βββ INDEX.md # Category overview
β βββ {items}.md # Individual entries
β
βββ sync/ # Optional: synced from built-in memory
βββ ...
The user defines the categories. Common examples:
projects/ β detailed project contextpeople/ β contact network with full contextdecisions/ β reasoning behind choicesknowledge/ β domain expertise, reference materialcollections/ β books, recipes, anything they collectSee memory-template.md for all templates.
Quick Reference
| Topic | File |
|-------|------|
| First-time setup | setup.md |
| All templates | memory-template.md |
| Organization patterns | patterns.md |
| Problems & fixes | troubleshooting.md |
Core Rules
1. Separate from Built-In Memory
This system lives in ~/memory/. Never modify:
memory/ folder (if it exists in workspace)Parallel, not replacement. Both systems work together.
2. User Defines Structure
During setup, ask what they want to store. Create categories based on their needs:
| They say... | Create |
|-------------|--------|
| "I have many projects" | ~/memory/projects/ |
| "I meet lots of people" | ~/memory/people/ |
| "I want to track decisions" | ~/memory/decisions/ |
| "I'm learning [topic]" | ~/memory/knowledge/[topic]/ |
| "I collect [things]" | ~/memory/collections/[things]/ |
No preset structure. Build what they need.
3. Every Category Has an Index
Each folder gets an INDEX.md that lists contents:
# Projects Index| Name | Status | Updated | File |
|------|--------|---------|------|
| Alpha | Active | 2026-02 | alpha.md |
| Beta | Paused | 2026-01 | beta.md |
Total: 2 active, 5 archived
Indices stay small (<100 entries). When full, split into subcategories.
4. Write Immediately
When user shares important information: 1. Write to appropriate file in ~/memory/ 2. Update the category INDEX.md 3. Then respond
Don't wait. Don't batch. Write immediately.
5. Search Then Navigate
To find information: 1. Ask first: "Is this in ~/memory/ or built-in memory?" 2. Search: grep or semantic search in ~/memory/ 3. Navigate: INDEX.md β category β specific file
# Quick search
grep -r "keyword" ~/memory/Navigate
cat ~/memory/INDEX.md # What categories exist?
cat ~/memory/projects/INDEX.md # What projects?
cat ~/memory/projects/alpha.md # Specific project
6. Sync from Built-In (Optional)
If user wants certain info copied from built-in memory:
~/memory/sync/
βββ preferences.md # Synced from built-in
βββ decisions.md # Synced from built-in
Sync is one-way: Built-in β this system. Never modify built-in.
7. Scale by Splitting
When a category grows large:
~/memory/projects/
βββ INDEX.md # "See active/, archived/"
βββ active/
β βββ INDEX.md # 30 active projects
β βββ ...
βββ archived/
βββ INDEX.md # 200 archived projects
βββ ...
What to Store Here (vs Built-In)
| Store HERE (~/memory/) | Keep in BUILT-IN | |------------------------|------------------| | Detailed project histories | Current project status | | Full contact profiles | Key contacts quick-ref | | All decision reasoning | Recent decisions | | Domain knowledge bases | Quick facts | | Collections, inventories | β | | Anything that grows large | Summaries |
Rule: Built-in for quick context. Here for depth and scale.
Finding Things
For Small Memory (<50 files)
# Grep is fast enough
grep -r "keyword" ~/memory/
For Large Memory (50+ files)
Navigate via indices:1. ~/memory/INDEX.md β find category
2. ~/memory/{category}/INDEX.md β find item
3. ~/memory/{category}/{item}.md β read details
For Huge Memory (500+ files)
Use semantic search if available, or hierarchical indices:~/memory/projects/INDEX.md β "web projects in web/"
~/memory/projects/web/INDEX.md β "alpha project"
~/memory/projects/web/alpha.md β details
Maintenance
Weekly (5 min)
Monthly (15 min)
When Memory is Slow
Common Traps
Security & Privacy
Data location:
~/memory/ on user's machineThis skill does NOT:
Related Skills
Install withclawhub install if user confirms:
decide - Decision tracking patternsescalate - When to involve humanslearn - Adaptive learningFeedback
clawhub star memoryclawhub syncβ‘ When to Use
User needs organized long-term storage beyond basic agent memory: detailed project histories, extensive contact networks, decision logs, domain knowledge, collections, or any structured data that grows over time.
βοΈ Configuration
On first use, read setup.md to configure the memory system with the user. Key decisions:
1. What categories do they need?
2. Should we sync anything from built-in memory?
3. How do they want to find things?