sandgardenhqOfficial·63 Agent Skills Included

sgai

Goal-driven multi-agent software factory with human supervision

Turns plain-English software goals into coordinated multi-agent workflows that plan, build, review, and test code. Replaces step-by-step prompting with a dashboard where a coordinator delegates work to specialist developer and reviewer agents. Enforces proof of completion through tests and linting, and extracts reusable skills from finished sessions so results improve over time.

130|8

↗

npx skills add sandgardenhq/sgai --all -g -y

Available:

Gives agents the repository's coding, testing, and verification rules, and documents each specialist agent's role so the coordinator can route work correctly.

All Skills in This Repository (63)

Pure Emerald Level Indicators

📦 In Repo

sandgardenhq

new-openai-sdk-app

Generate OpenAI Agents SDK applications for Python or TypeScript agents.

test-driven-development

Write failing tests before production code using a red-green-refactor cycle.

testing-anti-patterns

Enforce TDD practices and prevent mock-related testing anti-patterns.

systematic-debugging

Locate and validate root causes before implementing fixes.

project-completion-verification

Summarize GOAL.md task completion by parsing markdown checkboxes.

markdown-navigation

Generate hierarchical outlines and extract sections from markdown files using bash utilities.

run-long-running-processes-in-tmux

Runs commands in detached tmux sessions with lifecycle management and monitoring.

dispatching-parallel-agents

Dispatch one focused agent per independent software test failure in parallel.

retrospective

Analyze completed AI factory session artifacts to identify inefficiencies and propose enhancements.

code-cleanup

Audit generated code against GOAL.md and log changes in PROJECT_MANAGEMENT.md.

finishing-a-development-branch

Run tests before offering merge, pull request, or cleanup actions.

browser-bug-testing-workflow

Automate browser debugging with tmux servers and Playwright interactions.

Official

Advanced

Frequently Asked Questions

FAQPage Schema

How to install sgai?▼

Run `npx skills add sandgardenhq/sgai --all -g -y` in your terminal to install all skills in this suite globally.

What does sgai do?▼

Sgai turns a written goal into a working software project: a coordinator agent plans the work, delegates to specialist developer and reviewer agents, and only marks the goal done when tests pass.

How do I run a project with sgai?▼

Create a GOAL.md describing the outcome, run `sgai serve`, and open the local dashboard to approve the plan and monitor progress.

Can I control sgai from Claude Code or other AI harnesses?▼

Yes. Sgai exposes an MCP endpoint and HTTP skills covering workspace management, session control, monitoring, and human interaction, so any MCP-capable harness can drive it.

Does sgai work without constant supervision?▼

Yes. Self-drive mode runs fully autonomously, while brainstorming mode pauses to ask you questions at key decision points before continuing.