Agent skill

cass-memory

Contextual learning system that remembers patterns and rules from past sessions. Use to get relevant context before tasks, record outcomes, and build a personal playbook of coding patterns.

View SKILL.md on GitHub Repository
Stars 21
Forks 5

Install this agent skill to your Project

npx add-skill https://github.com/johnlindquist/claude/tree/main/skills/cass-memory

SKILL.md

CASS Memory - Contextual Learning System

Build and use a personal playbook of coding patterns learned from your sessions.

Prerequisites

The cm CLI should be available (part of cass-memory system).

Initialize:

bash
cm init
# Or with a starter playbook
cm init --starter typescript
cm init --starter react
cm init --starter python
cm init --starter go

CLI Reference

Get Context for a Task

bash
# THE main command - get relevant rules before starting work
cm context "Description of your task" --json

This returns:

  • Relevant rules from your playbook
  • Anti-patterns to avoid
  • History snippets from similar past work

Reflection (Extract Patterns)

bash
# Run reflection on recent sessions
cm reflect --json

# Specify lookback period
cm reflect --days 7 --json
cm reflect --days 30 --json

Playbook Management

bash
# List all rules
cm playbook list --json

# Get specific rule details
cm playbook get b-8f3a2c --json

# Add a new rule
cm playbook add "Always use optional chaining for nested object access" --json

Feedback on Rules

bash
# Mark rule as helpful
cm mark b-8f3a2c --helpful --json
cm mark b-8f3a2c --helpful --reason "Prevented null error" --json

# Mark rule as harmful
cm mark b-8f3a2c --harmful --json
cm mark b-8f3a2c --harmful --reason "Caused false positive" --json

Record Session Outcomes

bash
# Record success
cm outcome --status success --json
cm outcome --status success --rules "b-8f3a2c,b-4d2e1f" --json

# Record failure
cm outcome --status failure --text "Build failed due to type error" --json

# Mixed results
cm outcome --status mixed --text "Partial completion" --json

Statistics

bash
# Get playbook stats
cm stats --json

Top Rules

bash
# Show most effective rules
cm top --json
cm top 5 --json
cm top 20 --json

Health Check

bash
# Check system health
cm doctor --json

# Auto-fix issues
cm doctor --fix --json

Find Stale Rules

bash
# Rules without recent feedback
cm stale --json
cm stale --days 30 --json
cm stale --days 60 --json

Validate Rules

bash
# Validate a proposed rule against history
cm validate "Proposed rule text" --json

Explain Rule Origin

bash
# Show evidence and reasoning for a rule
cm why b-8f3a2c --json

Usage Statistics

bash
cm usage --json

Starter Playbooks

bash
# List available starters
cm starters --json

Workflow Patterns

Session Start

bash
# Get context before starting a task
cm context "Implement user authentication with JWT" --json

During Work

When a rule helps:

bash
cm mark b-8f3a2c --helpful --json

When a rule leads astray:

bash
cm mark b-8f3a2c --harmful --reason "Not applicable to this framework" --json

Session End

bash
# Record outcome
cm outcome --status success --rules "b-8f3a2c,b-4d2e1f" --json

Periodic Maintenance

bash
# Weekly: Run reflection to extract new patterns
cm reflect --days 7 --json

# Monthly: Review stale rules
cm stale --days 30 --json

# Check system health
cm doctor --json

Building Your Playbook

bash
# Manually add a pattern you've learned
cm playbook add "Use React.memo() for components receiving complex objects as props" --json

# After adding, use it in context queries
cm context "Create a list component with filtering" --json

Rule Lifecycle

  1. Creation - Rules emerge from reflection or manual addition
  2. Usage - Rules surface in context queries
  3. Feedback - Mark as helpful/harmful based on experience
  4. Evolution - High-feedback rules rise, low-feedback rules become stale
  5. Retirement - Stale rules get reviewed and pruned

Best Practices

  1. Always get context first - Run cm context "task" before starting work
  2. Provide feedback - Mark rules as helpful/harmful
  3. Record outcomes - Track session success/failure
  4. Run reflection regularly - Weekly reflection extracts new patterns
  5. Review stale rules - Don't let old rules accumulate
  6. Add rules manually - When you learn something important

Integration Tips

Pre-Task Context

Before any significant coding task:

bash
CONTEXT=$(cm context "Your task description" --json)
# Use context to inform your approach

Post-Session Recording

At end of coding session:

bash
cm outcome --status success --text "Completed feature X" --json

Recommended Agent Skills

Expand your agent's capabilities with these related and highly-rated skills.

johnlindquist/claude

testgen

Generate tests using AI and run test suites. Use for generating unit tests, running coverage reports, and mutation testing.

21 5
Explore
johnlindquist/claude

article

Generate technical articles and documentation using AI. Use for writing blog posts, documentation, and technical content.

21 5
Explore
johnlindquist/claude

packx

Bundle code context for AI. ALWAYS use --limit 49k unless user explicitly requests otherwise. Use for creating shareable code bundles and preparing context for LLMs.

21 5
Explore
johnlindquist/claude

long-agent

Manage long-running agent sessions. Use for tracking progress in extended tasks, maintaining context across long sessions, and managing multi-step workflows.

21 5
Explore
johnlindquist/claude

db

Database operations for SQLite, PostgreSQL, and MySQL. Use for queries, schema inspection, migrations, and AI-assisted query generation.

21 5
Explore
johnlindquist/claude

investigate

Debug and investigate code issues using search and AI analysis. Use when stuck on bugs, tracing execution flow, or understanding complex code.

21 5
Explore

Didn't find tool you were looking for?

Be as detailed as possible for better results