OpenGemini SaaS Builder

Build SaaS projects end-to-end with OpenClaw as orchestrator and Gemini CLI as a local planning and generation copilot. Use when the user asks to brainstorm,...

Security Scan

The skill's requests, scripts, and runtime instructions are coherent with its stated purpose of orchestrating SaaS projects with Gemini, GitHub CLI, and Vercel; it does not request unrelated credentials or perform unexplained actions.

Purpose & Capability

The name and description match the actual behavior: the skill coordinates Gemini CLI for generation, gh for repo creation, and vercel for deployment. The included scripts call gemini, gh, and vercel as expected. There are no unrelated environment variables, binaries, or external services requested.

Instruction Scope

SKILL.md and the scripts limit operations to planning, docs generation, scaffolding, repo creation (when explicitly requested), and optional Vercel deployment. The preflight and safety rules explicitly require user confirmation before external side effects (creating repos, pushing, deploying). The scripts operate on local files and call standard CLIs; they do not read unexpected system paths or hidden credentials.

Install Mechanism

This is an instruction-only skill with bundled helper scripts and no install spec or remote downloads. No archives or external installers are pulled; the scripts rely on already-installed tools (gemini, gh, vercel), which is proportional to the purpose.

Credentials

The skill declares no required environment variables or credentials. The scripts check for existing CLI authentication (gh auth status, vercel whoami) but do not request or embed secrets. This level of access is appropriate for repo creation and deployment tooling.

Persistence & Privilege

always:false and normal autonomous invocation are used (platform default). The skill can perform external side effects (repo creation, push, deploy) but SKILL.md mandates explicit user confirmation before doing so. Users should be aware that if the agent is allowed to act autonomously and misinterprets intent, side effects could occur—this is a general platform consideration, not a fault of the skill itself.

Assessment

This skill appears to do what it says: generate docs, scaffold files, create GitHub repos (via gh), and deploy to Vercel (via vercel) when you explicitly request those actions. Before installing or invoking: 1) ensure you have and trust the gh, vercel, and gemini CLIs on your machine and that their authenticated accounts are the ones you want used; the scripts will call those CLIs and may push your current repository. 2) Never include secrets or sensitive credentials in prompts or allow the skill to auto-generate .env files with real credentials—review .env.example and any generated files before use. 3) Confirm external side effects explicitly when prompted (repo creation/push/deploy). 4) If you plan to allow autonomous invocation broadly, be aware an agent could attempt actions using installed CLIs if it misinterprets user intent; require explicit confirmation for destructive or external actions. If you want more assurance, review the helper scripts locally (they are small and readable) and test in a disposable repo before using on important projects.

OpenClaw SaaS Builder

Use this skill when the user wants full project orchestration instead of a quick snippet.

Core idea

OpenClaw is the driver. Gemini CLI is the brainstorming and generation copilot. GitHub CLI is for repo creation and GitHub operations. Vercel CLI is optional for deployment.

Do not treat Gemini as an unsupervised executor. Use it to generate plans, alternatives, code drafts, copy, architecture ideas, and reviews. OpenClaw remains responsible for file edits, shell execution, validation, and external side effects.

What this skill should produce

For a serious SaaS or web project, aim to create at least:

• requirements.md
• architecture.md
• Rules.md
• README.md
• TODO.md
• skills-plan.md
• source code scaffold or implementation

For deployment-ready work, also aim for:

• .env.example
• deployment.md
• CI/deploy notes

Preflight first

Before promising repo creation or deployment, run the preflight helper:

./skills/openclaw-saas-builder/scripts/preflight_saas_builder.sh

Interpretation:

• If Gemini CLI fails, continue with OpenClaw-only planning or ask whether to proceed without Gemini.
• If gh auth status fails, do not promise automatic repo creation yet.
• If vercel whoami fails, do not promise deployment yet.

If Vercel is required, the user should already have the CLI installed and authenticated. Typical checks:

which vercel
vercel --version
vercel whoami

Build modes

Read references/saas-build-modes.md when deciding scope. Use one of these modes explicitly:

• idea mode
• MVP mode
• production mode
• launch mode

Default workflow

Phase 1: Discovery and brainstorming

Clarify:

• product goal
• target users
• pain point
• MVP scope
• must-have features
• non-goals
• monetization or business purpose
• data and privacy concerns
• preferred stack or constraints

Use Gemini CLI to accelerate this phase. Prefer structured output. Good pattern:

• ask Gemini for a concise options list
• ask Gemini for tradeoffs
• ask Gemini for a JSON spec draft

Phase 2: Requirements package

Create these docs early:

• requirements.md
• architecture.md
• Rules.md
• skills-plan.md
• TODO.md

Use the bundled helpers when useful:

./skills/openclaw-saas-builder/scripts/generate_requirements.sh <project-name> '<brief>'
./skills/openclaw-saas-builder/scripts/generate_architecture.sh <project-name> '<brief>'
./skills/openclaw-saas-builder/scripts/generate_rules.sh <project-name> '<brief>'

Read references/saas-doc-templates.md for the expected sections.

If the user asks for diagrams, generate Mermaid diagrams in markdown first.

Phase 3: Implementation plan

Break work into small, ordered milestones. Use TODO.md or a milestone list. Good milestone pattern:

1. bootstrap repo
2. scaffold app
3. core domain logic
4. auth and data
5. UI and polish
6. tests
7. docs
8. repo publish
9. deploy

Phase 4: Build

Choose the build path based on complexity.

For small or medium tasks

• inspect files directly
• use Gemini CLI for spec, code drafts, or review
• edit files with OpenClaw tools
• run tests locally

For larger coding tasks

Prefer ACP or coding-agent for implementation loops. Use Gemini CLI as planning and review support, not as the only coding engine.

Phase 5: GitHub repository creation

Only do this when the user has explicitly asked for it or clearly requested end-to-end setup.

Typical local flow:

1. initialize git if needed
2. create README and baseline docs
3. verify gh auth status
4. create repo with a sensible name
5. set remote
6. commit initial version
7. push default branch

Use the helper when appropriate:

./skills/openclaw-saas-builder/scripts/create_github_repo.sh <repo-name> [--private|--public]

Use the GitHub skill for detailed command patterns.

Naming guidance:

• use a short, readable slug
• prefer product name or function over vague names
• avoid joke names unless the user asks

Phase 6: Optional Vercel deployment

Only do this when the user asked for deployment.

Before deploy, verify:

which vercel
vercel --version
vercel whoami

If Vercel auth is not ready, stop and say exactly that. Do not fake deployment readiness.

Use the helper when appropriate:

./skills/openclaw-saas-builder/scripts/deploy_vercel.sh

Typical flow:

1. verify framework and build output
2. confirm env vars
3. run Vercel link or init if needed
4. deploy
5. capture deployment URL
6. write deploy notes to deployment.md

Safety and approval rules

These actions change the outside world and should be treated as explicit-user-intent actions:

• creating remote repos
• pushing commits to remote
• creating cloud projects
• deploying to Vercel
• sending webhooks or emails

Planning, docs, scaffolding, and local implementation can proceed normally. For external side effects, confirm intent if the request is ambiguous.

How to use Gemini CLI inside this skill

Use gemini-cli skill patterns. Recommended helpers:

• ./skills/gemini-cli/scripts/check_gemini.sh
• ./skills/gemini-cli/scripts/gemini_json_validate.sh
• ./skills/gemini-cli/scripts/gemini_review.sh
• ./skills/gemini-cli/scripts/gemini_with_model.sh

Recommended Gemini uses:

• brainstorming product directions
• generating JSON specs
• drafting architecture alternatives
• drafting landing-page copy
• reviewing rewritten files
• naming ideas for repos or products

Bad use:

• blindly executing Gemini-generated shell
• trusting invalid JSON
• letting Gemini decide external side effects without review

Recommended outputs by project stage

Early stage

• requirements.md
• architecture.md
• Rules.md
• TODO.md

Mid stage

• code scaffold
• component and module plan
• .env.example
• test checklist

Late stage

• README.md
• deployment.md
• GitHub repo
• deploy URL

References

Read these when needed:

• references/saas-doc-templates.md
• references/repo-and-deploy-checklist.md
• references/saas-build-modes.md

Use skills/gemini-cli/references/website-prompts.md when the project includes a marketing site or landing page.

Publish readiness notes

This skill is meant to be practical, not magical. It assumes:

• Gemini CLI is installed and authenticated
• GitHub CLI is installed and authenticated for repo creation
• Vercel CLI is installed and authenticated for deployment
• OpenClaw remains the orchestrator for validation, editing, and safety decisions