grovv stack

grovv stack

 ██████╗ ██████╗  ██████╗ ██╗   ██╗██╗   ██╗
██╔════╝ ██╔══██╗██╔═══██╗██║   ██║██║   ██║
██║  ███╗██████╔╝██║   ██║██║   ██║██║   ██║
██║   ██║██╔══██╗██║   ██║╚██╗ ██╔╝╚██╗ ██╔╝
╚██████╔╝██║  ██║╚██████╔╝ ╚████╔╝  ╚████╔╝
 ╚═════╝ ╚═╝  ╚═╝ ╚═════╝   ╚═══╝    ╚═══╝

gro\/\/ stack — Production-First Project Scaffolding

A prompt-driven scaffolding system that an AI agent uses to generate production-ready codebases (new projects) or layer documentation, skills, conventions, and an agent team onto existing ones. It is conversation-driven and ask-first: it understands the product, users, constraints, and stack before generating anything.

This repository is not an application. Its output is documents, configuration, and an agent team in your project — never code in this repo.

---

Install

/plugin marketplace add constructyourself/grovv-stack
/plugin install grovv-stack@grovv

---

Kick it off

In any repository — new or existing:

Entry point	How
Command	/grovv — auto-detects new vs existing. Force the mode with /grovv new or /grovv adopt.
Natural language	"build out this project with grovv stack", "adopt grovv stack in this repo" — triggers the grovv-scaffold skill.

Both run the same workflow defined in grovv-stack-scaffold.md. For an existing project, the agent assesses the codebase and proposes an adoption plan before changing anything — it never overwrites working code without approval.

Working from a clone of this repo (without installing) also works: the agents and skills under .claude/ load as project-scope components.

---

What it generates

Into the target project:

product-spec.md            # What, who, why
development-plan.md         # Engineering plan
tech-spec.md               # Complete technical specification
docs/skills/               # 12-15 development best-practice guides
docs/prompts/              # Prompt specs (skills-builder, team-design, tech-spec, readme)
docs/architecture/         # Architecture decision records
.claude/agents/            # Project-specific agent team (additive to the six defaults)
.claude/skills/            # Skills the agents use + an orchestrator
README.md                  # Project README

The pipeline runs: structure + config → product spec → development plan → tech spec → prompt docs → skills-builder → team-design (harness) → linear-tracking → readme. The team-design step uses the bundled harness meta-skill to design a project-specific agent team, additive to the six grovv defaults. The linear-tracking step uses the Linear MCP to create a project and seed issues from the development plan.

---

What ships in the plugin

Component	Path
Command	commands/grovv.md
Kickoff skill	.claude/skills/grovv-scaffold/
harness meta-skill (Apache-2.0, vendored)	.claude/skills/harness/
Default agents	.claude/agents/ (scaffold, frontend, backend, testing, database, code-review)
Master directive	grovv-stack-scaffold.md

---

Non-negotiables

• Ask which frontend framework (Astro + React or Next.js) before any frontend code.
• Ask what Playwright should test before writing any E2E test.
• Production-first, security by default, zero data loss (transactions for multi-step data ops).
• Never overwrite working code in existing projects without an approved adoption plan.

---

License

grovv stack is licensed under the MIT License. Vendored third-party components keep their own licenses — the bundled harness meta-skill is Apache-2.0; see .claude/skills/harness/ATTRIBUTION.md.

---

gro\/\/ stack — Production-First Project Scaffolding