Use this guide to understand the local tools that make agentflow-sdlc work well. Start with read-only validation and then choose what to install yourself. The framework must propose installation options, not execute them automatically.
Run environment validation from this repository or from an installed copy:
node bin/cli.mjs doctor-env --target /path/to/project
node bin/cli.mjs doctor-env --target /path/to/project --jsondoctor-env is read-only. It reports mutated: false, lists found and missing tools, explains why each tool matters, and prints installation options. It does not install packages, edit shell profiles, authenticate GitHub, or change project files.
| Tool | Why it matters | Validation |
|---|---|---|
git |
Branches, commits, PR workflow, release tags, and framework sync all assume Git. | git --version |
| Node.js | Runs the CLI, validators, hooks, tests, and framework scripts. | node --version |
Use a current LTS version of Node.js. This repository declares its runtime expectation in package.json and uses Node-based scripts for cross-platform compatibility.
| Tool | Why it matters | Validation |
|---|---|---|
gh |
Optimized GitHub issue/PR/release workflow: create issues, post comments, open PRs, inspect checks, close integrated issues, and publish releases. | gh --version |
| Package manager | Runs project validation. This repo uses pnpm; adopting projects may use npm, pnpm, yarn, bun, or another configured tool. |
Project-specific |
gh is not always required for reading files or editing code, but the optimized workflow assumes it for durable GitHub evidence and lifecycle automation.
Supported direct routing clients are project-configurable:
claudecodexagypi
Optional meta-harness/runtime:
omnigent— use when the project wants Omnigent to supervise agents, policies, sandboxes, or collaboration.
agent-workflow.config.json can define routing.agents.<slug>.availabilityCommand. doctor-env reads those commands and reports whether the configured optional agent/runtime is available. Missing optional tools do not block the default single-agent workflow unless your project has chosen to require them.
qa-expert is an optional exploratory QA sidecar role. It is not part of the deterministic SDLC phase sequence, but its environment should be explicit when a project enables exploratory/manual QA.
Baseline tools:
| Tool | Why it matters | Validation |
|---|---|---|
gh |
Creates exploratory QA session issues, records findings, links follow-up bugs, and applies labels such as needs-test. |
gh --version |
| Vibium | Default qa-expert browser automation skill/tool for exploratory browser use and evidence capture. |
vibium --version or project-specific |
| Browser(s) | Manual UX flow validation, accessibility checks, responsive behavior, and browser-specific regressions. | Project-specific |
| Test data / seeded environment | Reproducible exploratory sessions without relying on private or production-only data. | Project-specific |
| Screenshot / screen recording tool | Captures evidence for visual bugs, reproduction steps, and QA handoffs. | Project-specific |
Default recommendation:
- Prefer Vibium for
qa-expertwhen browser-based exploratory QA is in scope. This is an opinionated default because it has been validated successfully for AI-assisted exploratory QA, but projects may override it when an existing QA stack is a better fit. - Vibium upstream repository: https://github.com/VibiumDev/vibium
- Vibium website/docs: https://vibium.com
Common project-specific additions:
- Playwright, Cypress, Selenium, or another browser automation runner for repro scripts;
- accessibility tooling such as Axe, Lighthouse, or browser devtools accessibility inspectors;
- API tools such as curl, HTTPie, Postman, or Bruno for negative-path API exploration;
- mobile simulators/emulators when the product has mobile acceptance paths;
- local services, seed scripts, fixture loaders, or sandbox credentials required for safe manual QA.
Document these in project conventions or agent-workflow.config.json as optional availability checks. doctor-env can report qa-expert dependencies when they are represented as configured optional tools; it should still remain read-only and only propose installation/setup options.
- GitHub integration lifecycle automation uses GitHub Actions plus
ghfor local investigation. The workflow itself runs in GitHub withcontents: read,pull-requests: read, andissues: writepermissions.
These are examples for humans/operators to choose from. Agents should present them for approval, not run them automatically.
# Git
brew install git
# GitHub CLI
brew install gh
# pnpm via Corepack
corepack enable pnpmFor Linux, prefer your distro package manager or each tool's official docs.
# Git
winget install --id Git.Git
# GitHub CLI
winget install --id GitHub.cliInstall Node.js from the official Node.js site or your preferred version manager.
- Git: https://git-scm.com/downloads
- Node.js: https://nodejs.org/
- pnpm: https://pnpm.io/installation
- GitHub CLI: https://cli.github.com/
- Omnigent: https://github.com/omnigent-ai/omnigent
When this document changes, keep lib/environment.mjs and doctor-env aligned. The docs should explain the same required/recommended/optional split that the command reports.