ScrubbyAI Convention checks for AI-assisted code.
Workflow FAQ Get the Beta
BETA · LIMITED ACCESS

Protect your codebase from AI slop.

AI assistants ship code faster, but move the review burden upstream.

Scrubby learns your codebase conventions and blocks inefficient, non-idiomatic AI code before it ships.

$ scrubby learn . --save repo-patterns.json → scanning codebase… ok → discovering conventions… ok → wrote "repo-patterns.json" $ git commit -m "Refactor billing" scrubby pre-commit: warning Non-idiomatic pattern in src/billing/ error Inefficient query helper (perf regression) Options: [f]ix patch [o]verride once [c]ancel $ scrubby f ✔ patch applied ✔ commit allowed

The paradox:

AI allows devs to ship more code faster, but it also increases review times and reduces overall codebase quality and consistency.

What changed

  • 40%+ of new code is AI-assisted (and rising)
  • Code compiles ≠ code belongs in your repo
  • Universal lint rules don’t capture team conventions

What it causes

  • Review becomes pattern-policing
  • More rework, churn, and rollbacks
  • Tech leads become full-time reviewers

Scrubby is a repo-specific quality gate

Scrubby learns your codebase conventions and enforces them automatically.

Learns your patterns

Scans your repo, discovers implicit conventions, and saves them locally for quick reference

Blocks slop early

Stops inefficient, non-idiomatic changes at pre-commit and PR stages, before review debt piles up.

Explains and fixes

AI assistant explains violations, suggests patches, and helps devs learn your team’s standards.

How it works

One onboarding pass, then continuous guardrails for every contributor.

1) Install → Setup wizard

Quick step-through configuration for your team's language and check preferences.

2) Learn the repo

Scrubby scans your repo and discovers your team's existing coding conventions.

3) Real-time linting

Inline violations appear as you code using dual-layer checks -- fast static + repo-specific AI.

4) Pre-commit hook

Scrubby's interactive commit hook checks staged files, flags violations, and suggests fixes.

5) PR quality gate

ScrubbyCI integration generates a pattern violation report and blocks merges until violation resolution.

6) Human- and CI-friendly outputs

Scrubby produces informative, historical logs in human-readable Markdown and structured JSON formats.

Multiple ways to work with Scrubby

Use Scrubby interactively while coding, or run it headlessly for automation and enforcement.

IDE Extension

  • Repo-contextual pattern discovery + repo save
  • Real-time linting + inline diagnostics
  • Chat assistant for explanations & fixes
Scrubby: Setup Wizard Scrubby: Learn Repo Patterns Scrubby: Explain Violation

CLI wrapper

  • Batch checks (staged, diff, or full repo)
  • Pre-commit + CI modes
  • Outputs: console / json / markdown
scrubby learn . --save repo-patterns.json scrubby check --staged scrubby ci --comment-pr --block-on error

FAQ

Short answers. No theater.

> Is Scrubby “just ESLint with AI”?

No. Linters enforce universal rules. Scrubby enforces your repo’s conventions by learning patterns from your codebase, then applying them consistently at pre-commit and PR.

> Will this slow down commits?

Scrubby is designed to be fast on the happy path. Pre-commit checks focus on staged files and use a dual-layer approach: quick static checks first, deeper AI checks only when needed.

> Do we need to use VS Code?

No. VS Code is the best interactive experience, but the CLI wrapper and CI integration let teams enforce the gate regardless of editor.

> What about privacy?

In beta, analysis uses an API model with a structured pipeline. The roadmap includes local/edge inference for zero latency and stronger privacy.

© Scrubby
Status: stable · quality: non-negotiable