line-cook

name	line-cook
description	AI-supervised development workflow. Use when running prep, cook, serve, tidy, or work commands, managing beads issues during sessions, or following prep→cook→serve→tidy cycle. Covers workflow orchestration, guardrails, and session management.

Line Cook

Structured AI workflow execution for disciplined development.

When to Use

• Starting a work session with "work"
• Running individual workflow steps: "mise", "prep", "cook", "serve", "tidy", "plate"
• Planning phases: "brainstorm", "scope", "finalize"
• Managing beads issues during execution
• Architecture decisions, audits, and quality checks
• Understanding workflow guardrails

Quick Reference

Command	Purpose
"brainstorm"	Explore problem space (divergent thinking)
"scope"	Structure work breakdown (convergent thinking)
"finalize"	Convert plan to beads and test specs
"mise" or "plan"	Full planning (brainstorm→scope→finalize)
"prep"	Sync git, show ready tasks
"cook"	Claim and execute a task with TDD cycle
"serve"	AI peer review of completed work
"tidy"	Commit, sync beads, push
"plate"	Validate completed feature
"run" or "work"	Full execution cycle
"decision"	Manage architecture decisions (ADRs)
"architecture-audit"	Audit codebase architecture
"plan-audit"	Audit planning quality
"help"	Show available commands
"loop"	Manage autonomous loop
"getting started"	Show workflow guide

Use @line-<command> syntax for explicit invocation (e.g., @line-brainstorm, @line-cook).

Core Workflow

Quick cycle (most common):

prep → cook → serve → tidy

Full service (feature delivery):

mise → prep → cook → serve → tidy → plate

Step-by-Step

1. Mise: Plan work breakdown (interactive)
2. Prep: Sync state, identify available work
3. Cook: Claim task, execute with guardrails, verify completion
4. Serve: AI reviews changes
5. Tidy: File discovered issues, commit, push
6. Plate: Validate feature completion (when applicable)

Guardrails

Line Cook enforces these disciplines:

• Sync before work - Always start with current state
• One task at a time - Focus prevents scope creep
• Verify before done - Tests pass, code compiles
• File, don't block - Discovered issues become new beads
• Push before stop - Work isn't done until it's pushed

TDD Cycle with Quality Gates

Cook phase follows Red-Green-Refactor with automatic quality checks:

RED: Write failing test

• Write test for the feature
• Verify test fails
• Taster agent reviews test quality
• Address critical issues before proceeding

GREEN: Implement minimal code

• Write simplest code to make test pass
• Verify tests pass
• No refactoring yet

REFACTOR: Clean up code

• Improve structure while tests pass
• Ensure tests still pass after refactoring

Quality Gates (must pass before completion):

• Tests pass (project-specific test command)
• Code builds (project-specific build command)
• Test quality approved by taster agent
• Code quality approved by sous-chef agent (serve phase)

Beads Integration

Line Cook uses beads for task management:

bd ready                              # Find unblocked tasks
bd update <id> --status in_progress   # Claim task
bd show <id>                          # Get task context
bd close <id>                         # Complete task
bd sync                               # Sync with git

Parking Lot Pattern

Tasks under "Retrospective" or "Backlog" epics are excluded from auto-selection. Explicit selection still works:

bd update <parked-task-id> --status in_progress

Error Handling

If a step fails:

• Prep fails: Fix sync issues, retry
• Cook fails: Continue to tidy to save progress
• Serve fails: Review skipped, continue to tidy
• Tidy fails: Create bead for follow-up
• Plate fails: Note incomplete, create follow-up bead

Reference

For full documentation, see:

• README.md in project root - Philosophy and installation
• steering/*.md files - Workflow instructions