Skip to content

Universal Skills

Jump to bottom
Varun Pratap Bhardwaj edited this page Apr 10, 2026 · 2 revisions

Universal Skills

SuperLocalMemory v2.1.0 includes 7 universal agent-skills that work as slash-commands in Claude Code, Continue.dev, and Cody. These ai-skills provide a consistent interface across different AI assistants while accessing the same local memory database.

Keywords: agent-skills, slash-commands, ai-skills, claude-code, continue-dev, cody

🎯 What Are Skills?

Skills are command-based interfaces for AI assistants that don't support MCP (Model Context Protocol) natively. They provide the same functionality as MCP tools but through a slash-command interface.

Key Benefits

  • Consistent Interface - Same commands work across multiple IDEs
  • No MCP Required - Works with any tool that supports commands
  • Local Execution - Everything runs on your machine
  • Single Database - Uses the same SQLite database as MCP and CLI

📦 The 7 Universal Skills

All skills use the slm-* prefix for consistency:

Skill Purpose Usage
slm-remember Save content to memory /slm-remember "content" --tags work,api
slm-recall Search memories /slm-recall "authentication"
slm-list-recent Display recent memories /slm-list-recent 10
slm-status System health and stats /slm-status
slm-build-graph Rebuild knowledge graph /slm-build-graph
slm-switch-profile Change active profile /slm-switch-profile work
slm-show-patterns Show learned patterns /slm-show-patterns

🚀 Installation

Automatic Installation

The install-skills.sh script automatically detects and configures skills for supported IDEs:

cd SuperLocalMemoryV2
./install-skills.sh

Auto-detects:

  • Claude Code (native skills)
  • Continue.dev (VS Code)
  • Cody (VS Code/JetBrains)

What it does:

  1. Detects installed IDEs
  2. Backs up existing configurations
  3. Merges skill definitions
  4. Creates necessary directories
  5. Sets correct permissions

Manual Installation

If you need to manually install skills:

Claude Code:

cp -r skills/* ~/.claude/skills/

Continue.dev:

# Edit .continue/config.yaml and add skills

Cody:

# Edit VS Code settings.json or JetBrains Cody config

📖 Skill Documentation

1. slm-remember

Save content to memory with automatic indexing.

Usage

# Basic
/slm-remember "We use FastAPI for all REST APIs"

# With tags
/slm-remember "JWT tokens expire after 24 hours" --tags security,auth,jwt

# With project
/slm-remember "Database uses PostgreSQL 15" --project myapp --tags database

# With importance (1-10)
/slm-remember "CRITICAL: Production deploy requires approval" --importance 10

Arguments

Argument Type Required Default Description
<content> string Yes - Text to remember
--tags string No None Comma-separated tags
--project string No "default" Project name
--importance integer No 5 Priority level (1-10)

What Happens

  1. Content Saved - Stored in SQLite database
  2. Entities Extracted - TF-IDF identifies key terms
  3. Graph Updated - Entities added to knowledge graph
  4. Patterns Learned - Analyzes for coding preferences
  5. ID Returned - Memory ID for future reference

Example Output

✓ Memory added with ID: 42
  Tags: security, auth, jwt
  Project: myapp
  Importance: 7
  Entities extracted: 3 (JWT, tokens, authentication)

2. slm-recall

Search memories using semantic search and full-text search.

Usage

# Basic search
/slm-recall "authentication"

# Search with limit
/slm-recall "database query" --limit 5

# Search by tag
/slm-recall "performance" --tags optimization

# Search by project
/slm-recall "api" --project myapp

Arguments

Argument Type Required Default Description
<query> string Yes - Search query
--limit integer No 10 Max results
--tags string No None Filter by tags
--project string No Current Filter by project

Search Methods

  1. Semantic Search - TF-IDF vector similarity
  2. Full-Text Search - SQLite FTS5 for exact matches
  3. Knowledge Graph - Related memories via clustering
  4. Pattern Context - Considers learned preferences

Example Output

Found 3 memories:

[1] ID: 42 (Score: 0.87)
    "Implemented JWT authentication with 24-hour expiry"
    Tags: security, auth, jwt
    Created: 2026-02-01
    Cluster: "Authentication & Security"

[2] ID: 15 (Score: 0.65)
    "OAuth2 integration for Google login"
    Tags: auth, oauth, security
    Created: 2026-01-28
    Cluster: "Authentication & Security"

[3] ID: 8 (Score: 0.52)
    "Added CSRF protection middleware"
    Tags: security, middleware
    Created: 2026-01-25
    Cluster: "Security Patterns"

3. slm-list-recent

Display recent memories with metadata.

Usage

# Show 10 most recent
/slm-list-recent

# Show 20 most recent
/slm-list-recent 20

# Filter by project
/slm-list-recent 10 --project work

Arguments

Argument Type Required Default Description
<limit> integer No 10 Number of memories
--project string No Current Filter by project

Example Output

Recent Memories (10):

[1] ID: 47 - 2 hours ago
    "Added Redis caching for API responses"
    Tags: performance, caching, redis

[2] ID: 46 - 5 hours ago
    "Fixed authentication bug in login flow"
    Tags: bugfix, auth

[3] ID: 45 - 1 day ago
    "Implemented rate limiting middleware"
    Tags: security, api

4. slm-status

System health, statistics, and diagnostics.

Usage

# Show all stats
/slm-status

# Brief status
/slm-status --brief

Example Output

╔══════════════════════════════════════════════════════════════╗
║  SuperLocalMemory v2.1.0 - System Status                     ║
╚══════════════════════════════════════════════════════════════╝

✓ Database: OK
  - Total memories: 47
  - Database size: 2.3 MB
  - Current profile: default
  - Profiles: 3 (default, work, personal)

✓ Knowledge Graph:
  - Clusters: 8
  - Total entities: 156
  - Average cluster size: 5.8 memories
  - Last build: 2 hours ago

✓ Pattern Learning:
  - Learned patterns: 12
  - High-confidence patterns: 5
  - Framework preferences: React (73%), FastAPI (68%)
  - Last update: 30 minutes ago

✓ Storage:
  - Tier 1 (full): 45 memories
  - Tier 2 (compressed): 2 memories
  - Tier 3 (archived): 0 memories
  - Space savings: 5%

System: Healthy

5. slm-build-graph

Rebuild knowledge graph with Leiden clustering.

Usage

# Build/rebuild graph
/slm-build-graph

# Build with custom resolution
/slm-build-graph --resolution 1.5

What It Does

  1. Entity Extraction - TF-IDF identifies important terms
  2. Similarity Calculation - Cosine similarity between memories
  3. Graph Construction - Create edges for similar memories
  4. Clustering - Leiden algorithm finds communities
  5. Auto-Naming - Generate descriptive cluster names

Example Output

Building knowledge graph...

✓ Processed 47 memories
✓ Extracted 156 entities
✓ Created 89 edges (similarity > 0.3)
✓ Detected 8 clusters:

Cluster 1: "Authentication & Security" (12 memories)
  - Top entities: JWT (8), OAuth (5), tokens (7), session (6)

Cluster 2: "Performance Optimization" (8 memories)
  - Top entities: caching (6), Redis (5), performance (8), speed (4)

Cluster 3: "React Components" (11 memories)
  - Top entities: React (11), components (9), hooks (7), state (6)

...

Graph build completed in 2.3 seconds.

6. slm-switch-profile

Change active memory profile for context isolation.

Usage

# Switch to work profile
/slm-switch-profile work

# Switch to personal profile
/slm-switch-profile personal

# Create and switch to new profile
/slm-switch-profile client-acme --create

Arguments

Argument Type Required Default Description
<profile> string Yes - Profile name
--create flag No False Create if doesn't exist

Profile Benefits

  • Isolated Contexts - Separate memories per profile
  • No Context Bleeding - Work/personal/client memories separate
  • Independent Graphs - Each profile has its own knowledge graph
  • Separate Patterns - Different learned preferences per profile

Example Output

Switching to profile: work

✓ Profile loaded: work
  - Total memories: 142
  - Knowledge clusters: 15
  - Learned patterns: 8
  - Last used: 3 hours ago

Profile 'work' is now active.

🔧 IDE-Specific Usage

Claude Code

Native skills support - Skills appear in command palette.

# Type / to see all skills
/slm-remember "content"
/slm-recall "query"

Features:

  • Auto-completion
  • Inline help
  • Syntax highlighting

Continue.dev (VS Code)

Slash commands - Skills work as custom commands.

Setup:

./install-skills.sh  # Auto-configures Continue

Usage:

# In Continue chat panel
/slm-remember "We use TypeScript for all frontend code"
/slm-recall "typescript configuration"

Features:

  • Tab completion
  • Command history
  • Integration with VS Code tasks

Cody (VS Code/JetBrains)

Custom commands - Skills as Cody commands.

Setup:

./install-skills.sh  # Auto-configures Cody

Usage:

# In Cody chat
/slm-remember "Added logging middleware"
/slm-status

Features:

  • Command suggestions
  • Context menu integration
  • JetBrains and VS Code support

🏗️ Skills Architecture

Metadata-First Design

Each skill includes a SKILL.md file with complete metadata:

---
name: slm-remember
description: Save content to SuperLocalMemory
version: "2.1.0"
license: AGPL-3.0
compatibility: "Requires SuperLocalMemory installed"
attribution:
  creator: Varun Pratap Bhardwaj
  role: Solution Architect & Original Creator
  project: SuperLocalMemory
---

Directory Structure

skills/
├── slm-remember/
│   ├── SKILL.md          # Metadata and documentation
│   ├── main.py           # Skill implementation
│   └── examples/         # Usage examples
├── slm-recall/
│   ├── SKILL.md
│   ├── main.py
│   └── examples/
└── ...

Benefits

  • Version Tracking - Each skill has semantic versioning
  • Attribution Preserved - Creator info in every skill
  • Self-Documenting - Complete docs in SKILL.md
  • IDE Agnostic - Same structure works everywhere

🔄 Skills vs MCP vs CLI

All three access methods use the same SQLite database:

Feature Skills MCP CLI
Interface Slash commands AI tools Terminal commands
Best For Claude/Continue/Cody Modern IDEs Scripts/Terminals
Requires IDE with commands MCP support Any shell
Setup install-skills.sh Auto in install.sh Auto in install.sh
Database ~/.claude-memory/memory.db Same Same

Example - Same Operation, Three Ways:

# Skills (Claude Code)
/slm-remember "We use FastAPI"

# MCP (Cursor)
"Remember that we use FastAPI"

# CLI (Terminal)
slm remember "We use FastAPI"

All three save to the same memory, queryable by all methods.

🐛 Troubleshooting

Skills Not Showing Up

Claude Code:

  1. Check ~/.claude/skills/ directory exists
  2. Verify skill directories are present
  3. Restart Claude Code
  4. Type / to see command list

Continue.dev:

  1. Check .continue/config.yaml has skills defined
  2. Reload VS Code window (Cmd+Shift+P → "Reload Window")
  3. Open Continue panel
  4. Type / to see commands

Cody:

  1. Check VS Code settings.json has cody.customCommands
  2. Reload VS Code
  3. Open Cody chat
  4. Type / for commands

"Skill Not Found" Error

Solution: Reinstall skills:

cd SuperLocalMemoryV2
./install-skills.sh --force

"Permission Denied"

Solution: Fix permissions:

chmod +x ~/.claude-memory/skills/*/main.py

Skills Return No Data

Solution: Check database:

slm status  # Should show memory count

If empty, add some memories first:

slm remember "Test memory"

📚 Full Skill Reference

For complete documentation of each skill, see:

  • skills/slm-remember/SKILL.md
  • skills/slm-recall/SKILL.md
  • skills/slm-list-recent/SKILL.md
  • skills/slm-status/SKILL.md
  • skills/slm-build-graph/SKILL.md
  • skills/slm-switch-profile/SKILL.md
  • skills/slm-show-patterns/SKILL.md

Or visit the GitHub repository.

🔗 Related Pages

Created by Varun Pratap Bhardwaj

SuperLocalMemory V3 — Your AI Finally Remembers You. 100% local. 100% private. 100% free.

Part of Qualixar | Created by Varun Pratap Bhardwaj | GitHub

SuperLocalMemory V3

Getting Started

Reference

Architecture

Enterprise

Release Notes

V2 Documentation

Clone this wiki locally