Skip to main content

Overview

The settings command provides interactive configuration of GSD workflow agents and model profile through a guided multi-question prompt. Configure which optional agents run and which Claude models are used.

Syntax

/gsd:settings

How It Works

  1. Ensures config exists - Creates ~/.claude/get-shit-done/config.json with defaults if missing
  2. Reads current settings - Loads and parses existing configuration
  3. Interactive 5-question prompt - Presents choices with current values pre-selected
  4. Merges config - Updates only changed settings
  5. Writes config file - Saves merged configuration
  6. Displays confirmation - Shows updated settings with quick command references

Configuration Options

1. Model Profile

Controls which Claude model each agent uses. Options:
  • quality - Claude Opus for all agents (highest quality, highest cost)
  • balanced - Sonnet for most, Opus for critical agents (recommended)
  • budget - Sonnet for all agents (lower cost, good quality)

2. Research Agent

Runs codebase research before planning phases. Options:
  • enabled - Research runs by default (recommended)
  • disabled - Skip research, faster planning
Affects: /gsd:plan-phase

3. Plan Checker

Validates plans before execution, suggests improvements. Options:
  • enabled - Plans reviewed before execution (recommended)
  • disabled - Execute plans without review
Affects: /gsd:execute, /gsd:quick --full

4. Verifier Agent

Validates implementation after execution. Options:
  • enabled - Implementation verified against plan (recommended)
  • disabled - Skip verification, faster execution
Affects: /gsd:execute, /gsd:quick --full

5. Git Branching

Controls branch creation during phase execution. Options:
  • enabled - Create feature branch per phase
  • disabled - Work directly on current branch (default)
Affects: /gsd:execute

Usage Examples

Initial configuration

/gsd:settings
Interactive prompt:
GSD Configuration

1. Model Profile
   Which model profile?
   → quality (Claude Opus for all agents)
     balanced (Sonnet + Opus for critical) [current]
     budget (Claude Sonnet for all agents)

2. Research Agent
   Run codebase research before planning?
   → enabled [current]
     disabled

3. Plan Checker
   Review plans before execution?
   → enabled [current]
     disabled

4. Verifier Agent
   Verify implementation after execution?
   → enabled [current]
     disabled

5. Git Branching
   Create feature branches during execution?
     enabled
   → disabled [current]

Press Enter to use current values, or select options:

Change to budget mode

/gsd:settings
You select:
1. budget
2. disabled
3. disabled  
4. disabled
5. disabled
Confirmation:
✓ Settings updated

Active Configuration:
- Model: budget (Sonnet for all agents)
- Research: disabled
- Plan checker: disabled
- Verifier: disabled
- Branching: disabled

Fastest mode - minimal agent overhead.
Good for: prototyping, simple codebases, low token budgets.

Quick toggles:
- /gsd:set-profile quality|balanced|budget
- /gsd:plan-phase --no-research
- /gsd:quick (always skips optional agents)

Quality mode for critical project

/gsd:settings
You select:
1. quality
2. enabled
3. enabled
4. enabled
5. enabled
Confirmation:
✓ Settings updated

Active Configuration:
- Model: quality (Opus for all agents)
- Research: enabled
- Plan checker: enabled
- Verifier: enabled
- Branching: enabled

Maximum quality mode - full agent workflow.
Good for: production systems, complex domains, critical code.

Note: Higher token usage. Monitor costs in usage dashboard.

Model Profiles

Quality Profile

{
  "planner": "claude-opus-4",
  "executor": "claude-opus-4",
  "verifier": "claude-opus-4",
  "plan_checker": "claude-opus-4",
  "research": "claude-opus-4",
  "debugger": "claude-opus-4"
}
Cost: ~3-5x vs balanced Quality: Highest reasoning, best for complex work
{
  "planner": "claude-opus-4",
  "executor": "claude-sonnet-4",
  "verifier": "claude-sonnet-4",
  "plan_checker": "claude-opus-4",
  "research": "claude-sonnet-4",
  "debugger": "claude-opus-4"
}
Cost: ~1.5-2x vs budget Quality: High quality where it matters, efficient elsewhere

Budget Profile

{
  "planner": "claude-sonnet-4",
  "executor": "claude-sonnet-4",
  "verifier": "claude-sonnet-4",
  "plan_checker": "claude-sonnet-4",
  "research": "claude-sonnet-4",
  "debugger": "claude-sonnet-4"
}
Cost: Baseline Quality: Good for most tasks, excellent value

Config File Location

~/.claude/get-shit-done/config.json
Format:
{
  "model": "balanced",
  "research": "enabled",
  "plan_check": "enabled",
  "verifier": "enabled",
  "branching": "disabled"
}

Default Settings

If no config exists, GSD uses:
{
  "model": "balanced",
  "research": "enabled",
  "plan_check": "enabled",
  "verifier": "enabled",
  "branching": "disabled"
}
Balanced quality and cost, all quality gates enabled.

Per-Command Overrides

Settings can be overridden per command:
# Skip research for this phase only
/gsd:plan-phase --no-research

# Full quality for quick task
/gsd:quick --full

# No verification for this execution
/gsd:execute 4 --no-verify

When to Change Settings

Use Quality Profile

  • Production systems
  • Complex business logic
  • Security-critical code
  • Regulatory compliance
  • High-stakes projects

Use Balanced Profile

  • Most projects (recommended default)
  • Balances cost and quality
  • Full-stack applications
  • Team development

Use Budget Profile

  • Learning projects
  • Prototyping
  • Simple CRUD apps
  • Personal projects
  • Token budget constraints

Quick Commands

Faster ways to change specific settings:
# Change model profile only
/gsd:set-profile quality
/gsd:set-profile balanced  
/gsd:set-profile budget

# Toggle research for one command
/gsd:plan-phase --no-research

# Full quality for one quick task
/gsd:quick --full
  • set-profile - Quick model profile switching
  • plan-phase - Uses research setting
  • execute - Uses verifier and branching settings
  • quick - Always skips optional agents (—full overrides)