Skills & Commands

Overview

CEMS provides skills (slash commands) that enable AI assistants to search and store persistent memories across sessions. Skills are available in Claude Code, Cursor, Codex, and Goose with slight variations based on IDE capabilities.

Available Skills

/recall - Search Memories

Search your memories for relevant information using semantic search. Usage:

/recall <search query>

Examples:

/recall What do I prefer for backend development?
/recall database conventions
/recall How do we handle authentication?
/recall deployment process

How It Works:

Detects the current project from git remote (e.g., “Chocksy/cems”)
Converts your query to embeddings using text-embedding-3-small
Searches both personal and shared memory (by default)
Boosts same-project results, penalizes cross-project noise
Ranks results by relevance score with time decay
Returns the most relevant matches

Options:

/recall --scope personal my coding preferences
/recall --scope shared team conventions
/recall --scope both authentication patterns
/recall --category decisions database choices
/recall --limit 10 all architecture decisions

MCP Tool Used: memory_search with:

query: Your search text (required)
scope: “personal”, “shared”, or “both” (default: both)
max_results: Maximum results (default: 10, range: 1-20)
max_tokens: Token budget for results (default: 4000)
enable_graph: Include related memories via graph traversal (default: true)
enable_query_synthesis: Expand query with LLM for better retrieval (default: true)
project: Auto-detected from git remote (org/repo format)
raw: Debug mode to bypass relevance filtering (default: false)

For Truncated Results: When search results are truncated (content ends with ...), fetch the full document:

{
  "tool": "memory_get",
  "arguments": {
    "memory_id": "the-memory-id-from-search-result"
  }
}

/remember - Add Personal Memory

Add a memory to your personal memory store. Usage:

/remember <fact or information to remember>

Examples:

/remember I prefer Python for backend development
/remember The database schema uses snake_case for column names
/remember User authentication uses JWT tokens with 24h expiry

How It Works: The memory_add MCP tool automatically:

Extracts atomic facts from your input
Checks for existing similar memories
Decides whether to ADD (new), UPDATE (modify existing), or skip (duplicate)
Stores with timestamp and access tracking

Options:

/remember --category preferences I like dark mode in all editors
/remember --category decisions We chose PostgreSQL for the main database
/remember --tags auth,security Session tokens expire after 1 hour of inactivity

MCP Tool Used: memory_add with:

content: Your input (required)
scope: “personal” (automatic for /remember)
category: Category for organization (default: “general”)
- Available: preferences, conventions, architecture, decisions, workflow, errors, learnings, general
tags: Array of tags for filtering
source_ref: Project reference (auto-detected as “project:org/repo”)

Project Context: When running in a git repository, always include the project reference:

Get git remote: git remote get-url origin
Extract org/repo from URL:
- SSH: [email protected]:org/repo.git → org/repo
- HTTPS: https://github.com/org/repo.git → org/repo
Pass as source_ref: "project:org/repo"

This enables project-scoped memory recall where memories are boosted when searching from the same project. Add a memory to the shared team memory store. Usage:

/share <fact or information to share with team>

Examples:

/share API endpoints follow REST conventions with versioning (/api/v1/...)
/share Deploy process: merge to main, wait for CI, then run deploy script
/share Architecture decision: microservices communicate via message queue

Use Cases:

Team conventions and standards
Architecture decisions (ADRs)
Deployment processes
Codebase patterns
Onboarding knowledge

Requirements: You must have CEMS_TEAM_ID configured for shared memory to work. Options:

/share --category architecture We use hexagonal architecture pattern
/share --category conventions All dates are stored in UTC
/share --tags deploy,process Deploy requires manual approval for production

MCP Tool Used: memory_add with:

content: Your input
scope: “shared”
category: Default “general” or specified
tags: As specified

/forget - Remove Memory

Remove a memory from the system. Usage:

/forget <memory_id>

Examples:

/forget abc123def456
/forget --hard abc123def456  # Permanent deletion

How It Works: By default, memories are archived (soft delete):

They won’t appear in searches
They can be recovered if needed
They’re preserved for audit purposes

Use --hard for permanent deletion when you want the memory completely removed. Finding Memory IDs: Memory IDs are returned when you:

Add a memory (/remember)
Search memories (/recall)
List memories via CLI

Example Workflow:

/recall old preferences
# Returns memories with IDs like "abc123..."

/forget abc123
# Archives that memory

MCP Tool Used: memory_forget with:

memory_id: The ID to remove (required)
hard: true for permanent deletion (default: false)

/context - Show Memory Status

Show the current memory system status and context. Usage:

/context

Output Shows:

Current user ID
Team ID (if configured)
Storage location
Memory counts by scope
Category breakdown
Scheduler status

Example Output:

CEMS Memory System Status
========================================

User ID: razvan
Team ID: engineering
Storage: /Users/razvan/.cems

Personal Memories:
- Total: 47
- By category:
  - preferences: 12
  - decisions: 8
  - patterns: 15
  - context: 7
  - general: 5

Shared Memories:
- Total: 23
- By category:
  - conventions: 10
  - architecture: 8
  - processes: 5

Scheduler: Running
- Next consolidation: Tonight 3:00 AM
- Next summarization: Sunday 4:00 AM
- Next reindex: Feb 1 5:00 AM

MCP Resources Used:

memory://status - System status
memory://personal/summary - Personal memory summary
memory://shared/summary - Shared memory summary

Foundation Guidelines Skill

View foundation guidelines (rules, principles, constraints) stored in CEMS memory. Usage: When invoked or when the user asks about project rules/guidelines:

Detect the current project from git remote

Search for foundation guidelines via MCP:

{
  "tool": "memory_search",
  "arguments": {
    "query": "foundation guidelines rules principles constitution",
    "max_results": 20,
    "scope": "both",
    "project": "org/repo"
  }
}

Filter results to only include memories that:
- Have tags containing “foundation” or “constitution”
- OR have category “guidelines”

Context: Foundation guidelines are stable rules that apply across all sessions:

Coding standards and conventions
Project-specific constraints
Team agreements and principles
Quality gates and review criteria

They differ from regular memories in that they are meant to be always-on constraints, not just recalled information.

IDE-Specific Implementations

Claude Code

Locations:

Skills: ~/.claude/skills/cems/
Commands: ~/.claude/commands/
6 hooks, 6 skills, 2 commands
MCP server config in ~/.claude.json

Available:

/recall - Search memories
/remember - Add personal memory
/share - Add team memory
/forget - Delete memory
/context - Show status
Memory guide skill for proactive usage

Cursor

Locations:

Skills: ~/.cursor/skills/
3 hooks, 5 skills
MCP config in ~/.cursor/mcp.json

Available:

cems-recall - Search memories
cems-remember - Add memory
cems-forget - Delete memory
cems-share - Share with team
cems-context - Memory status

Codex

Locations:

Commands: ~/.codex/commands/
Skills: ~/.codex/skills/
3 commands, 2 skills
MCP config in ~/.codex/config.toml

Available:

recall command - Search memories
remember command - Add memory
foundation command - View guidelines
recall skill - Search memories
remember skill - Add memory

Goose

Location:

MCP extension config in ~/.config/goose/config.yaml

Available: All MCP tools via the cems extension:

memory_search
memory_add
memory_forget
memory_update
memory_get
memory_maintenance

Search Tips

Always pass project: Auto-detect from git remote to filter cross-project noise
Use natural language: “how do we handle authentication” > “auth”
Be specific: “Python backend database conventions” > “conventions”
The system uses semantic matching, not just keywords
Results include relevance scores and time decay ranking
Use raw: true to debug retrieval when results seem wrong

What to Store

Good candidates for memory:

User preferences and style choices
Project conventions and naming patterns
Architecture and infrastructure decisions
Debugging insights and solutions to recurring problems
Workflow patterns and deployment processes

What NOT to store:

Session-specific context (current task, temporary state)
Information already in the codebase
Build output or error messages being debugged right now
Speculative conclusions from a single observation

CLI Usage - Command-line interface reference
Memory Management - Memory lifecycle and best practices
Observer Daemon - Background session learning

Get Started

Core Concepts

IDE Integration

Using CEMS

Server Deployment

Advanced

Skills & Commands

Overview

Available Skills

/recall - Search Memories

/remember - Add Personal Memory

/forget - Remove Memory

/context - Show Memory Status

Foundation Guidelines Skill

IDE-Specific Implementations

Claude Code

Cursor

Codex

Goose

Search Tips

What to Store

Build docs developers (and LLMs) love

Get Started

Core Concepts

IDE Integration

Using CEMS

Server Deployment

Advanced

​Overview

​Available Skills

​/recall - Search Memories

​/remember - Add Personal Memory

​/share - Add Team Memory

​/forget - Remove Memory

​/context - Show Memory Status

​Foundation Guidelines Skill

​IDE-Specific Implementations

​Claude Code

​Cursor

​Codex

​Goose

​Search Tips

​What to Store

​Related Pages

Build docs developers (and LLMs) love

Overview

Available Skills

/recall - Search Memories

/remember - Add Personal Memory

/share - Add Team Memory

/forget - Remove Memory

/context - Show Memory Status

Foundation Guidelines Skill

IDE-Specific Implementations

Claude Code

Cursor

Codex

Goose

Search Tips

What to Store

Related Pages