Eazyr

eazyr

Turn "clone it and good luck" into a five-minute path from clone to a running app.

You produce an onboarding kit that lives in the repo:

The one rule

Never write a step you have not verified against the repo. Invented steps are worse than absent ones — they cost the reader the time it takes to discover they're fake.

This ships with a scanner so you don't have to rely on memory. eazyr scan --json returns detected facts with the file each one came from. Prefer it over reading twenty files by hand, and cite its provenance rather than guessing.

If a fact is missing, write > **TODO:** and say what you'd need. Never fill the hole with something plausible.

Workflow

1. Scan

bashCopy code
npx eazyr scan --json

Returns: ecosystems (stack, package manager, version ranges, real commands from the manifest), env.vars (each with required, secret, and sources: ["file:line"]), compose and services, ports, verify (health endpoint or test command), steps (the ordered setup path), and gaps (what's missing).

Read the human summary first if you just want orientation:

bashCopy code
npx eazyr scan

The scanner is a strong first pass, not an oracle. It is regex-based, skips node_modules, and caps at 4000 files. Verify anything surprising by opening the file it cites. If ecosystems is empty or truncated is true, fall back to reading the repo yourself — references/detect.md is the manual lookup table.

2. Generate the first draft

bashCopy code
npx eazyr init          # add --force to overwrite, --dry-run to preview

Writes all four files from the scan. Existing files are skipped unless --force, so this is safe to run in a repo that already has some docs.

This is a draft, not a deliverable. It knows the facts; it does not know your project.

3. Edit it into something true

This is the part only you can do. Work through the generated files and:

• Resolve every TODO. The generator emits one wherever it couldn't determine a fact.
• Read the sources. For each sources: ["src/config.ts:14"], open that line and write what the variable actually does. "Stripe API key" beats a repeated variable name.
• Name where secrets come from. STRIPE_SECRET_KEY → "Create a test-mode key at dashboard.stripe.com/apikeys". The generator cannot know this; a reader is stuck without it.
• Check CI. .github/workflows/ is the setup that provably works. If it does something the quickstart doesn't, the quickstart is wrong.
• Fix the ordering if the project has a real constraint the scanner missed — a seed step that must precede first boot, a service that needs a warm-up.
• Cut what isn't on the critical path. Under 120 lines. Everything else is README material.

references/quickstart.md has the section-by-section spec, references/doctor.md the check conventions, references/detect.md the manual detection table.

4. Verify

bashCopy code
npx eazyr doctor        # runs the derived checks against this machine, writes nothing
bash scripts/doctor.sh  # and confirm the committed script agrees

Then run the install and start commands if you can. Best effort, in this order:

1. doctor passes, and fails informatively when you unset a required variable.
2. The install step completes.
3. The app starts and the verify step returns what the quickstart claims.

Report honestly: what you ran, what passed, what you could not check. If you couldn't run the app, say so plainly rather than implying the path is proven.

When the scanner comes up empty

Unrecognised stack, or a repo too unusual to detect? Fall back to the manual workflow — the artifacts are still the goal:

1. Read CI, then the task runner, then the container files. CI has to pass, so it doesn't lie.
2. Reconstruct: prerequisites → clone → install → configure → services → run → verify.
3. Copy assets/doctor.sh and assets/doctor.ps1 into scripts/, and edit only the CHECKS block — the runner is complete.
4. Write QUICKSTART.md from assets/QUICKSTART.template.md.

Scope

Write the kit and the scripts that support it. Don't refactor the project, change its dependencies, or rewrite the main README — QUICKSTART.md links from the README, it doesn't replace it. Never copy a real value out of an existing .env; use a placeholder and say you did.

Files

• references/detect.md — manual stack detection, per ecosystem
• references/quickstart.md — QUICKSTART.md section spec
• references/doctor.md — check catalog and script conventions
• assets/QUICKSTART.template.md, assets/doctor.sh, assets/doctor.ps1, assets/env.example.template — the templates init fills in