Quality checker for software teams

See the result

The same code, with a visible standard

The checker does not stop at a red number: it shows what was disorganized and how it looks when every decision has a place. This is the output a person and an AI agent can both use.

src/services/users.ts

header maintained by ArduMaker Checker

scan ready

View comparison

Before · disorganized

After · organized

4 findings

no blockers

No clear types, no documented contract and mixed responsibilities.

Explicit contract, edge validation and an output that is quick to review.

structure

everything in its place

handoff

less context lost

reading

10 lines, one intent

The change

Stop relying on someone remembering.

Before

• Everyone organizes code their own way.
• Structural issues appear once they affect the team.
• Review depends on memory, judgment and available time.
• An AI agent can produce correct but disorganized changes.

After

• The same standards apply to every repository.
• Hooks review work before it reaches the commit.
• Every finding says what it found, expected and where it is.
• JSON output lets agents and automations integrate without ambiguity.

What it checks

198 rules grouped by type of work

The catalog comes from our internal standards and is public so any team can adopt the same quality format. This is not a list of tips: every rule has an identifier, severity and verifiable result.

• Code and names: structure, signatures, complexity and conventions.
• Security: secrets, credentials, sensitive logs and unsafe configuration.
• Repository: README, Git, root files, names and required documents.
• Configuration: Django, Vite, JavaScript, infrastructure and deployments.
• AI work: stable JSON, explicit context and agent-ready output.
• Formatting: headers, lines, columns, whitespace and consistent files.

How it works

Automatic control, human judgment

The checker walks each file once, measures concrete values and returns exit codes useful for a hook or deploy. It only updates the headers maintained by the tool; everything else is fixed on purpose.

• Install it per repository with `ardumaker init` and check every commit.
• Findings distinguish errors, warnings and disabled rules.
• Activate a family or a specific rule with `--only` and `--skip`.
• Your ArduMaker account authorizes the terminal without anonymous identity.

Exceptions you can understand

A useful standard does not pretend every project is identical. Configuration lets you tune thresholds, change severities and document the team's decision.

• Rules can be activated by family or ID.
• Rules disabled by decision explain why they do not run by default.
• Thresholds come from the rule itself so documentation and behavior stay aligned.

What it does not promise

A checker does not replace design, tests or human review. It can flag structure and discipline, but it cannot decide whether a product matters or whether an architecture is right.

Adopting it has an initial cost: there will be findings to address. The benefit appears when the repository stops accumulating the same problem in every delivery.