smartlisa

Lisa

Read the blog post: Why I Built Lisa

Lisa keeps it organized. The Ralphs get it done.

Meet Lisa. The smart AI project manager that never loses context. She breaks down complex projects into Milestones → Epics → Stories and hands off tasks — to Claude, your dev team, Linear, or a swarm of Ralph Wiggum loops running overnight.

One planner. Any executor. Whether it's a senior engineer, a junior dev, or an AI agent bumping into walls until tests pass — Lisa doesn't care who picks up the card. She just needs it done.

The new bottleneck isn't coding. AI writes code faster than teams can plan and review it. Lisa helps PMs, tech leads, and solo devs break work into well-scoped stories before the Ralphs start running — so you're not scrambling to keep up with your own agents.

Installation

CLI Tool

npm install -g smartlisa

          _     ___ ____    _
         | |   |_ _/ ___|  / \
         | |    | |\___ \ / _ \
         | |___ | | ___) / ___ \
         |_____|___|____/_/   \_\

Claude Code Skills

Lisa works as a Claude Code skill. Choose one installation method:

Via Plugin (Recommended)

# Add the marketplace
/plugin marketplace add klosowsk/lisa

# Install the plugin
/plugin install lisa@lisa

As Project Skills (anyone cloning repo gets them)

# Copy to your project
cp -r path/to/lisa/skills/lisa your-project/.claude/skills/
cp -r path/to/lisa/skills/lisa-work your-project/.claude/skills/

As Personal Skills (available in all your projects)

# Symlink to your personal skills directory
ln -s $(pwd)/skills/lisa ~/.claude/skills/lisa
ln -s $(pwd)/skills/lisa-work ~/.claude/skills/lisa-work

Using with Claude Code

Lisa is a Claude Code skill. Just ask:

"Help me plan this feature with Lisa"

"/lisa status"

"Work on the next story"

Claude uses Lisa's structured plans to systematically implement your project — no more wandering in circles.

CLI Commands

lisa status              # Project overview
lisa status board        # Kanban view of stories
lisa status show E1.S2   # Details for a specific item

lisa discover init       # Start a new project
lisa discover            # Continue discovery

lisa plan milestones        # View roadmap
lisa plan add-epic          # Add a feature
lisa plan generate-stories  # Create stories from PRD

lisa feedback mark E1.S2 done   # Update status
lisa validate                   # Check plan integrity

LLM as Runtime

Lisa treats the AI as an operating system. Every command returns structured context that both humans and agents can read — same interface, same data, different consumer.

$ lisa status board --milestone M1

                    ═══ Lisa Board ═══

TODO          IN_PROGRESS   REVIEW        DONE
─────────────────────────────────────────────────
E1.S3         E1.S2                       E1.S1
E2.S1
E2.S2

$ lisa status show E1.S2

═══ Story: E1.S2 ═══
Add login form

  Epic: Authentication
  Status: in_progress
  Type: feature

Description
  Create login form with email/password fields and validation

Acceptance Criteria
  [ ] Email field with format validation
  [ ] Password field with show/hide toggle
  [ ] "Remember me" checkbox persists session
  [ ] Error messages display below fields
  [ ] Loading spinner during submission

Architecture Context (from architecture.md)
  ## Authentication Flow
  - JWT tokens stored in httpOnly cookies
  - Refresh token rotation on each request
  ...

Context resets. Agents don't remember. Lisa does. Any agent can pick up where the last one left off — just run lisa status and get full context.

How It Works

Project
└── Milestone (M1)              # "MVP", "Beta", "Launch"
    └── Epic (E1)               # "Auth", "Payments"
        ├── prd.md              # Full requirements
        ├── architecture.md     # Technical design
        └── Story (E1.S1)       # "Add login form"

Each epic carries a full PRD and architecture doc — the complete context an agent needs to make good decisions. No more "I don't have enough context" excuses.

Stories flow: todo → in_progress → review → done

All data lives in .lisa/ — version it, share it, let any agent pick up where the last one left off.

License

MIT