line

Audit codebase architecture, code smells, and quality metrics

Explore problem space before planning (divergent thinking)

Validate completed epic and create acceptance documentation

Select and execute a task with completion guardrails

Record, list, or supersede architecture decisions (ADRs)

Diagnose issues

Convert plan to beads and create test specifications (execution)

Learn the workflow

Contextual help for Line Cook commands

Verify setup

Manage autonomous loop execution from TUI

Create work breakdown before starting implementation (orchestrates brainstorm→sample→scope→finalize)

Interactive walkthrough

Audit bead/plan structure, dependencies, and quality with auto-fix

Validate completed feature and create acceptance documentation

Sync state, load context, show ready tasks

Run full workflow cycle (prep→cook→serve→tidy)

Walk through user experience before scoping (experiential thinking)

Create structured work breakdown (convergent thinking)

Review changes via headless Claude and file issues

Commit changes, sync beads, and push to remote

Recent changes

Reviews epic-level E2E and smoke test coverage - validates user journeys and cross-feature integration before epic completion

Automated issue triage agent triggered by GitHub Actions. Analyzes new issues, searches the codebase for context, classifies type, applies labels, and responds with structured analysis. Not invoked directly — runs via the issue-agent workflow.

Reviews feature acceptance and BDD test quality - verifies acceptance criteria coverage and Given-When-Then structure

Simplify and refine recently modified code for clarity, consistency, and maintainability. Use this agent after sous-chef review to polish code and apply auto-fixable findings. This is an action agent that edits code. Examples: <example> Context: Sous-chef review returned APPROVED with auto-fixable findings. assistant: "Now let me polish the code and apply the auto-fixable review findings." <Task tool call to launch polisher agent with list of modified files and auto-fixable findings> </example> <example> Context: Sous-chef review returned APPROVED with no auto-fixable findings. assistant: "Let me polish these changes for clarity before tidy." <Task tool call to launch polisher agent> </example>

Use this agent when a logical chunk of code has been written or modified and needs review before proceeding. This includes after implementing a feature, fixing a bug, refactoring code, or completing any task that involves code changes. The agent should be called proactively after significant code changes to catch issues early. Examples: <example> Context: User asked to implement a new feature and the code has been written. user: "Add a function to validate email addresses" assistant: "Here is the email validation function I've implemented:" <function implementation completed> assistant: "Now let me use the sous-chef agent to review this code for correctness, security, and style before we proceed." <Task tool call to launch sous-chef agent> </example> <example> Context: User asked to fix a bug and the fix has been applied. user: "Fix the null pointer exception in the user service" assistant: "I've identified and fixed the null pointer exception by adding proper null checks:" <bug fix completed> assistant: "Let me have the sous-chef agent review this fix to ensure it's complete and doesn't introduce new issues." <Task tool call to launch sous-chef agent> </example> <example> Context: User completed a refactoring task. user: "Refactor the authentication module to use the new token service" assistant: "I've refactored the authentication module. Here are the changes:" <refactoring completed> assistant: "Before we consider this complete, I'll use the sous-chef agent to review the refactored code for any issues." <Task tool call to launch sous-chef agent> </example>

Reviews test quality during TDD cycle - ensures tests are isolated, fast, repeatable, clear with proper structure