Project Settings

Project Settings let you update the core configuration for a project. Open them by clicking the gear icon next to the project name in the sidebar. This navigates to /settings/project/:id.

Settings Reference

Name

The display name for the project shown in the sidebar and on the welcome screen card. Changing the name does not affect the repository or any thread directories.

Setup Script

A shell command that runs automatically in a new thread’s directory immediately after the thread is created (i.e. after the clone completes).

Examples:

npm install
bun install
pip install -r requirements.txt
cargo build
bundle install

The script runs in the thread’s working directory. If it fails, the thread is still created but a warning is shown. Leave this blank if no initial setup is needed.

Validation Script

An optional shell command that runs automatically after each agent response completes. Use this to verify the agent’s work — for example, running linters, type checks, or test suites.

If the validation script fails, you see a warning in the UI. The agent’s changes are not reverted automatically; use the validation output to decide whether to ask the agent to fix issues.

Examples:

npm run lint
bun run typecheck
pytest

Run Command

The command used by the Run/Stop button in the toolbar to start your development server. Click the Suggest button to auto-detect a dev script from package.json.

Examples:

bun dev
npm run dev
python manage.py runserver
rails server

Each thread has its own independent instance of this process. See Dev Server for details on starting, stopping, and viewing output.

Commit Method

Controls how the Git Ops button handles finished work:

Method	Behaviour
pull-request	After committing and pushing, Tarsk creates a pull request on GitHub using the gh CLI. Recommended for team projects where changes go through review.
direct	Commits are pushed directly to the thread’s branch. Suitable for personal projects or when you handle PRs yourself.

The gh CLI must be installed and authenticated for the pull-request method. See cli.github.com for setup instructions.

Command Approval

Controls whether bash commands and skill scripts require explicit approval before running:

Setting	Behaviour
Allow all commands	Commands run without per-command prompts (default)
Require approval	Each bash command and skill script shows an approval dialog

When approval is required, you can define allowed command patterns — commands matching these patterns (e.g. npm test, git status) run automatically without prompting.

You can also toggle Confirm commands temporarily from the chat mode selector. See Permissions for details.

Environment Variables

Add key-value pairs that are injected into the agent’s shell environment for all threads in this project. Values are stored encrypted locally.

• Click Add variable to create a new entry
• Variable names are normalized to uppercase
• Existing values are masked; leave the value field blank to keep the current value
• Remove variables with the X button

Environment variables are available to bash commands, skill scripts, and MCP servers configured for the project.

Open With

The preferred application used when clicking Open With in the toolbar. Choose from:

• VS Code
• Cursor
• Windsurf
• Xcode
• Android Studio
• Kiro
• Terminal
• Devin Desktop
• Antigravity

This preference is stored per-project and applies to all threads.

Custom System Prompt

By default, Tarsk auto-generates the agent’s system prompt from project context (AGENTS.md, rules, skills). You can override this with a custom system prompt:

1. Open Project Settings.
2. Write your custom system prompt in the System Prompt field.
3. Toggle Use Custom System Prompt to enable it.

When enabled, the agent uses your custom prompt instead of the auto-generated one. This gives you full control over how the agent behaves in this project, but you lose the automatic inclusion of AGENTS.md, rules, and skills context. Use this when precise control over the agent’s instructions matters.

Tools

Under Advanced → Tools, choose which built-in agent tools are available in this project. Each tool has a checkbox; uncheck a tool to disable it.

Disabled tools are hidden from the agent in every thread for this project. The agent cannot call them during conversations. All tools are enabled by default.

Disabling MCP blocks every MCP server tool for the project. See Chat for the full tool list.

Deleting a Project

To delete a project, click Delete Project at the bottom of the settings page. A confirmation dialog lists all threads that will be removed. Deletion removes the project record and all thread directories from disk. This cannot be undone.