From smith
This skill should be used when the user asks to configure Smith, set project preferences, customize agent behavior, or when any Smith command or agent needs to read project-local settings. It should also be used when the user asks "how do I configure Smith" or wants to change output paths.
How this skill is triggered — by the user, by Claude, or both
Slash command
/smith:configThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Smith supports per-project configuration via a `.smith.local.yaml` file in the project root. This file is user-managed, not committed to git, and provides project-specific settings that Smith agents and commands read at runtime.
Smith supports per-project configuration via a .smith.local.yaml file in the project root. This file is user-managed, not committed to git, and provides project-specific settings that Smith agents and commands read at runtime.
project-root/
└── .smith.local.yaml
The file lives in the project root, not inside .claude/ or docs/.
This file should be added to .gitignore:
.smith.local.yaml
Remind the user to add this entry when creating the file for the first time.
YAML
# Smith project configuration
# Directory where Smith writes research, plans, and reports.
# Default: docs/
output_directory: docs/
# Automatically critique and revise plans after initial generation.
# Default: false
plan_auto_critique: false
# Run git history analysis after research.
# Default: true
research_history: true
All fields are optional. An empty file or a file with only some fields is valid.
| Field | Type | Default | Description |
|---|---|---|---|
output_directory | string | docs/ | Base directory where Smith writes documents. Research files go to <output_directory>/research/, plans to <output_directory>/plans/, and reports to <output_directory>/reports/. |
plan_auto_critique | boolean | false | When true, the /plan command automatically runs one round of critique (Critic agent) and revision (Planner agent) after producing the initial plan. |
research_history | boolean | true | When true, the /research command runs the Historian agent after the Researcher to append git history analysis. Set to false to skip the history phase. |
The schema is intentionally open. Users may add additional keys as needed; agents should read what they recognize and ignore the rest.
If .smith.local.yaml does not exist, or a field is absent, use the defaults listed in the schema. Configuration is always optional; its absence must never block execution.
When a user asks to configure Smith or set up project settings:
.smith.local.yaml to the project root..smith.local.yaml is already in .gitignore. If not, remind the user to add it.npx claudepluginhub brettbuddin/claude-plugins --plugin smithAnalyzes a project context and generates Smith 2.0 agent definition files. Selects STATE/ACT classes and produces instance agents. Auto-triggered on 'create agent' requests.
Scans existing codebase and interviews you to bootstrap SpecKit projects with CLAUDE.md, constitution.md, and full .specify/ scaffolding. Supports presets for Flask-React, FastAPI-Next, CLI-Python, and Express-React stacks.
Configures Plan-Build-Run settings: workflow depth, model profiles, features, git branching/mode, and gates. Interactive or direct args like 'depth standard' or 'feature auto_continue on'.