Skill Commands

This is the complete reference for Lineup's skill commands. For a conceptual overview, see Skills.

Overview

Lineup workflows are generated from canonical templates in .lineup-core/skills/ and rendered to host-specific SKILL.md files.

Workflow	Claude command	Codex command	OpenCode command	Generated files (install-time)
Kick-off	/lineup:kick-off	$lineup-kick-off	/lineup-kick-off	skills/kick-off/SKILL.md, .agents/skills/lineup-kick-off/SKILL.md, ~/.config/opencode/skills/lineup-kick-off/SKILL.md
Configure	/lineup:configure	$lineup-configure	/lineup-configure	skills/configure/SKILL.md, .agents/skills/lineup-configure/SKILL.md, ~/.config/opencode/skills/lineup-configure/SKILL.md
Explain	/lineup:explain	$lineup-explain	/lineup-explain	skills/explain/SKILL.md, .agents/skills/lineup-explain/SKILL.md, ~/.config/opencode/skills/lineup-explain/SKILL.md
Playbook	/lineup:playbook	$lineup-playbook	/lineup-playbook	skills/playbook/SKILL.md, .agents/skills/lineup-playbook/SKILL.md, ~/.config/opencode/skills/lineup-playbook/SKILL.md
Digest	/lineup:digest	$lineup-digest	/lineup-digest	skills/digest/SKILL.md, .agents/skills/lineup-digest/SKILL.md, ~/.config/opencode/skills/lineup-digest/SKILL.md

Do not edit generated skill files directly. Edit .lineup-core/skills/**; host files are generated during lineup install and validated in CI with ./dev check.

Kick-off (/lineup:kick-off / $lineup-kick-off / /lineup-kick-off)

The main entry point for all Lineup workflows.

Syntax

/lineup:kick-off [task-description | tactic-name] [--from-stage N]
$lineup-kick-off [task-description | tactic-name] [--from-stage N]
/lineup-kick-off [task-description | tactic-name] [--from-stage N]

Arguments

Argument	Required	Description
task-description	No	Free-text description of the work to do
tactic-name	No	Name of a tactic to execute (kebab-case, matches a .yaml filename)
--from-stage N	No	Restart the pipeline from stage N, loading cached outputs for stages 0 through N-1

If no argument is provided, the skill enters menu mode.

When --from-stage N is used, the orchestrator loads cached stage outputs from .lineup/.cache/ for all stages before N and begins execution at stage N. If any required cache file is missing, the command reports which stage is missing and suggests a lower restart point.

Behavior

With a task description: The orchestrator evaluates the description, selects a pipeline tier based on complexity, and runs the appropriate stages.

/lineup:kick-off Refactor the authentication module to use JWT tokens
$lineup-kick-off Refactor the authentication module to use JWT tokens
/lineup-kick-off Refactor the authentication module to use JWT tokens

With a tactic name: The orchestrator looks for the named tactic in .lineup/tactics/ (project) and the plugin's tactics/ directory (built-in). If found, it runs the tactic's stage sequence. If not found, it reports the error and lists available tactics.

/lineup:kick-off brownfield-docs
$lineup-kick-off brownfield-docs
/lineup-kick-off brownfield-docs

With no arguments (menu mode): If tactics exist (project or built-in), the orchestrator presents a selection menu showing each tactic's name and description, plus options for the default pipeline and custom input. If no tactics exist, it prompts for a task description.

/lineup:kick-off
$lineup-kick-off
/lineup-kick-off

Initialization

Before starting pipeline stages or tactic execution, kick-off runs an initialization sequence defined in the host init file (skills/kick-off/INIT.md on Claude, .agents/skills/lineup-kick-off/INIT.md on Codex, ~/.config/opencode/skills/lineup-kick-off/INIT.md on OpenCode):

1. Agent configuration overrides -- reads user override files from ~/.claude/lineup/agents/, validates them, and merges with plugin defaults
   • Codex path: ~/.codex/lineup/agents/
   • OpenCode path: ~/.config/opencode/lineup/agents/
2. Memory migration -- one-time migration of global agent memory to project-scoped memory (skipped silently if already done)
3. Tactic resolution -- discovers tactics from .lineup/tactics/ and the plugin's tactics/ directory, presents selection if available

This initialization runs on every kick-off invocation, before any stages execute.

Pipeline stages

When running the default pipeline, kick-off executes Stage 0 (Triage) followed by up to 7 stages:

Stage	Name	Agent	User interaction
0	Triage	Orchestrator (no spawn)	None -- runs invisibly
1	Clarify	Orchestrator	Answers structured questions
2	Research	Researcher	None (wait for findings)
3	Clarification Gate	Orchestrator	Answers follow-up questions
4	Plan	Architect	Must approve before implementation
5	Implement	Developer	None (wait for changes)
6	Verify	Reviewer	Reviews the report
7	Document (optional)	Documenter	Opts in or skips

Triage (Stage 0) is a lightweight orchestrator-only analysis that classifies complexity, identifies affected areas, and produces search targets. It does not count toward the "Stage N/7" progress labels. Stages 1-7 may be skipped depending on the selected tier. See Pipeline Tiers.

Tactic execution

When a tactic is selected, its stage list replaces the default pipeline entirely. The orchestrator:

1. Prompts for variable values (if the tactic defines variables)
2. Executes each stage in order, respecting optional and gate controls
3. Passes context between stages (upstream output feeds downstream)
4. Presents verification criteria after all stages complete

Stage labels use the tactic's count: a 3-stage tactic shows "Stage 1/3", "Stage 2/3", "Stage 3/3".

Rules

• The orchestrator never implements code itself -- it delegates to the developer agent
• The orchestrator never does deep exploration -- it delegates to the researcher agent
• User approval is always required before moving from Plan to Implement
• Structured multiple-choice prompts are used for all user decisions in Clarify, Clarification Gate, and Document stages (AskUserQuestion on Claude, native question tool on OpenCode)

Configure (/lineup:configure / $lineup-configure / /lineup-configure)

Interactive agent configurator.

Syntax

/lineup:configure
$lineup-configure
/lineup-configure

Arguments

None. The skill is fully interactive.

Behavior

The configurator walks through five steps:

Step	What happens
1. Read	Reads all agent files and displays current frontmatter as a summary table
2. Ask	Presents configuration categories (model, tools, memory, reset) and collects choices
3. Preview	Shows the final frontmatter for each agent that will change
4. Apply	Writes override YAML files to host override directory (~/.claude/lineup/agents/, ~/.codex/lineup/agents/, or ~/.config/opencode/lineup/agents/)
5. Confirm	Reports what changed (which agents, which fields, old and new values)

Configuration categories

Category	Options
Model	Keep defaults, set one for all agents, set per-agent
Tools	Replace tools, add tools, remove tools, no changes
Memory	Keep defaults, set one for all agents, set per-agent
Ollama	Enable Ollama for research, disable Ollama, no changes
Reset	Restore all agents to factory defaults

What it modifies

Writes override files to host override directories (~/.claude/lineup/agents/ for Claude, ~/.codex/lineup/agents/ for Codex, ~/.config/opencode/lineup/agents/ for OpenCode) containing only the fields you changed (model, tools, memory). Never modifies agent .md files.

Validation

• Models must be haiku, sonnet, or opus
• Memory must be user, project, or local
• Tools are comma-space separated strings
• A preview is always shown before changes are written
• User confirmation is required before applying

Default values (used by Reset)

Agent	Model	Memory	Tools
researcher	haiku	project	Read, Grep, Glob, LS, WebFetch, WebSearch
architect	opus	project	Read, Grep, Glob, LS, Write
developer	opus	project	Read, Grep, Glob, LS, Edit, Write, Bash, NotebookEdit
reviewer	opus	project	Read, Grep, Glob, LS, Bash
documenter	opus	project	Read, Grep, Glob, LS, Write, WebFetch
teacher	opus	project	Read, Grep, Glob, LS, WebFetch, WebSearch

Explain (/lineup:explain / $lineup-explain / /lineup-explain)

Structured codebase explanation via the built-in explain tactic.

Syntax

/lineup:explain <question>
$lineup-explain <question>
/lineup-explain <question>

Arguments

Argument	Required	Description
question	Yes	A question about a project component, pattern, or decision

Behavior

This skill is an alias. It invokes kick-off explain mode (/lineup:kick-off explain on Claude, $lineup-kick-off explain on Codex, /lineup-kick-off explain on OpenCode) with the user's question as the task description. The kick-off workflow resolves the built-in explain tactic and executes its two stages:

1. Research: A researcher agent explores the codebase, focusing on the topic the user asked about
2. Explain: A teacher agent produces a structured explanation based on the research findings

Output

A YAML document following the templates/teacher.yaml schema, containing:

• Learning objectives
• Prerequisites
• Structured explanation with code examples from the actual codebase
• Suggestions for further exploration

The output is ephemeral -- it appears in the conversation and is not written to any file.

What questions work best

Questions about specific components, patterns, data flows, and architectural decisions produce the best results. See Use Explain for examples of effective questions.

Override

Create .lineup/tactics/explain.yaml in your project to override the built-in explain tactic with a custom workflow.

Playbook (/lineup:playbook / $lineup-playbook / /lineup-playbook)

Interactive tactic management wizard.

Syntax

/lineup:playbook
$lineup-playbook
/lineup-playbook

Arguments

None. The skill is fully interactive.

Behavior

The playbook skill walks through up to eight steps depending on the mode selected:

Step	What happens
1. Discover	Reads tactic schema, example templates, project tactics, and built-in tactics
2. Mode	Presents mode selection: Create, Import, Edit, Delete
3. Name/desc	Collects tactic name (kebab-case) and description
4. Stage builder	Guides through building an ordered list of stages
5. Verification	Collects human-readable verification criteria
6. Variables	Defines variables with cross-reference validation
7. Validate	Runs schema and semantic validation, shows YAML preview
8. Write	Writes the tactic to .lineup/tactics/<name>.yaml

Steps 3-7 are skipped or adapted depending on the mode.

Modes

Mode	Steps used	Description
Create	3, 4, 5, 6, 7, 8	Build a new tactic from scratch
Import	Selection + optional 3-7, 8	Copy an example template, optionally customize
Edit	Selection + targeted 3-7, 8	Modify an existing project tactic
Delete	Selection, confirmation	Remove a project tactic

Stage types and conventional agents

The stage builder presents these type-agent pairings as defaults:

Type	Default agent
clarify	researcher
research	researcher
clarification-gate	architect
plan	architect
implement	developer
verify	reviewer
document	documenter
explain	teacher

Any agent can be paired with any stage type -- these are conventions, not constraints.

Common stage patterns

The stage builder offers five pre-built patterns as starting points:

Pattern	Stages
Research-first	research, plan, implement, verify
Quick fix	plan, implement, verify
Documentation pass	research, plan, document
Full pipeline with controls	research (optional), plan (gate: approval), implement, verify, document (optional)
Investigation only	research, explain

Validation

Schema validation (blocks write):

• Name is kebab-case, 3-50 characters, matches filename
• Description is non-empty
• At least one stage exists
• Stage types and agent names are valid
• Variable names are snake_case
• All ${variable_name} references resolve to defined variables

Semantic validation (warns only):

• verify stage without verification criteria
• Verification criteria without a verify stage
• implement stage without a preceding plan stage
• plan stage without gate: approval
• Unused defined variables
• implement as the first stage

What it modifies

Writes, renames, or deletes files in .lineup/tactics/ only. Never modifies repository example files (examples/tactics/) or built-in tactics (tactics/).

Rules

• Preview and explicit confirmation are required before every write
• Example tactics and built-in tactics are read-only
• Rename in edit mode writes the new file before deleting the old one
• YAML formatting matches the example tactics exactly (header comments, pipe-blocks, field order)
• Structured multiple-choice prompts are used for all decisions (AskUserQuestion on Claude, native question tool on OpenCode)

Digest (/lineup:digest / $lineup-digest / /lineup-digest)

Standalone codebase overview generator.

Syntax

/lineup:digest [output-path]
$lineup-digest [output-path]
/lineup-digest [output-path]

Arguments

Argument	Required	Description
output-path	No	File path for the generated digest (default: DIGEST.md at project root)

Behavior

The digest skill runs a lightweight three-phase pipeline:

Phase	Agent	What happens
1. Research	3× Researcher (parallel)	Scan directory structure, entry points, modules, deps, tests, config
2. Structure	Architect	Organize findings into a coherent outline
3. Write	Documenter	Write the final markdown file

No approval gates or clarification stages — the pipeline runs end-to-end without user interaction.

Output

A markdown file (default DIGEST.md) containing:

• Project overview and tech stack
• Architecture and directory structure
• Module breakdown with file references
• Data flow description
• Key patterns and conventions
• Getting started guide
• Contributing guidelines (when discoverable)

The file begins with an auto-generated header and can be regenerated by running the skill again.

Ollama integration

When Ollama is enabled (via /lineup:configure), researcher agents use mcp__ollama__ollama_chat and mcp__ollama__ollama_generate to summarize large files and modules, reducing token cost.

Initialization

The skill runs its own lightweight initialization:

1. Agent configuration overrides -- reads user override files from ~/.claude/lineup/agents/ (Claude), ~/.codex/lineup/agents/ (Codex), or ~/.config/opencode/lineup/agents/ (OpenCode)
2. Ollama detection -- checks for Ollama config and MCP server availability