A Claude Code plugin for AI-assisted project management workflows. WIP guides your development from initial idea through shipped feature, using specialized agents and structured commands.

• Claude Code installed
• beads (bd) CLI installed

1. Add the marketplace:

   /plugin marketplace add github:clows/wip-marketplace
   

2. Install the plugin:

   /plugin install wip
   

3. Verify installation:

   /wip:test
   

1. Initialize beads in your project directory:

   bd init

   This creates a .beads/ directory for issue tracking.

2. Create your first task:

   bd create "My first feature" -t feature

   Or use the command: /wip:task:create

3. Check available work:

   bd ready

When you start a Claude Code session, Claude will greet you with your current work status (active tasks, ready work, etc.).

1. Initialize beads (if not already done):

   bd init

2. Import existing work (optional):

   • Create issues for in-flight work: bd create "Existing feature" -t feature
   • Set priorities: bd update <id> --priority 1

3. Start using WIP commands - they work immediately after plugin installation.

WIP follows a structured development flow:

brainstorm → plan → implement → test → review → ship

Each phase has a dedicated command that sets up the right context and spawns specialized agents.

1. Start with an idea:

/wip:workflow:brainstorm

Explore requirements, capture ideas, define scope. Creates tasks in beads.

2. Plan the implementation:

/wip:workflow:plan

Design architecture, break down into subtasks, identify dependencies.

3. Implement a task:

/wip:workflow:implement

Focus on a single task. The implementer agent writes code with limited scope.

4. Add tests:

/wip:workflow:test

Generate and run tests for your implementation.

5. Review changes:

/wip:workflow:review

Check code quality, security, and adherence to project patterns.

6. Ship it:

/wip:workflow:ship

Prepare documentation, create commits, and generate release notes.

Check what's ready to work on:

bd ready

See current progress:

/wip:task:status

Break a large feature into subtasks:

/wip:task:breakdown

When you start a Claude Code session, WIP injects context about your project into Claude's awareness. Claude will then greet you with:

• Current in-progress work
• Tasks that are ready (no blockers)
• Project statistics

This helps you pick up where you left off without running commands.

Claude automatically uses these skills when relevant:

You don't invoke skills directly—Claude decides when they're helpful.

Commands spawn focused agents with limited tools and context:

Agents run in separate contexts, keeping responses targeted and efficient.

WIP uses beads for issue tracking. Key commands:

bd ready              # Show unblocked work
bd list               # All issues
bd create "Title"     # New issue
bd update <id> --status in_progress  # Claim work
bd close <id>         # Complete work
bd show <id>          # Issue details

Issues are stored in .beads/issues.jsonl and version-controlled with your code.

• Start small: Use /wip:workflow:brainstorm for new ideas, even small ones
• Check status often: Run bd ready or /wip:task:status to stay oriented
• Let agents focus: Each agent has a specific job—trust the workflow phases
• Commit with issues: The .beads/ directory should be committed with code changes

"bd: command not found" Install beads: pip install beads-cli or see beads installation

No tasks showing at session start Run bd init in your project, then create some tasks with bd create.

Commands not found after install Restart Claude Code or run /plugin list to verify WIP is installed.

MIT