quality

One quality workflow for polyglot repositories

I built quality for repositories where “run the checks” stops being a simple instruction.

A modern product may contain Rust services, a Swift application, Android modules, TypeScript packages, and an Astro site. Each ecosystem already has good analyzers. The missing layer is a predictable way to detect them, run them together, normalize the results, and explain what a repository expects.

quality is that layer. It coordinates native tools rather than replacing them.

Goals

• Preserve ecosystem expertise by invoking the analyzers teams already trust.
• Give every repository one memorable workflow for checking, formatting, fixing, diagnosing, and adopting policy.
• Make gradual adoption practical with explicit configuration and baselines for existing findings.
• Use the same policy locally and in CI with structured output for people, agents, and GitHub.

What I built

• A native Rust CLI with init, doctor, check, format, fix, baseline, repositories, instructions, completions, and CI-generation commands.
• Thirteen built-in adapters covering Cargo fmt, Clippy, SwiftLint, SwiftFormat, Android Lint, detekt, ktlint, ESLint, Astro Check, Prettier, CSpell, Knip, and Actionlint.
• Repository task discovery that preserves canonical package scripts and monorepo-specific type-check or validation semantics.
• Changed-file execution that sends relevant paths to file-capable tools while escalating configuration changes to full checks.
• Baseline support that records existing diagnostics and fails only for new regressions without hiding missing tools or analyzer crashes.
• A GitHub Action with verified downloads, pull-request annotations, job summaries, SARIF reports, and configurable failure thresholds.

Technical highlights

• Concurrent execution: independent checks run together by default, with --fail-fast available when the first failure matters more than a complete report.
• Normalized diagnostics: pretty, JSON, SARIF, and GitHub output formats turn different analyzer conventions into one result model.
• Custom adapters: teams can add organization-specific tools with explicit commands, file modes, extensions, configuration files, and parsers.
• Cross-repository operations: audit a folder of repositories, preview missing policy, and apply configuration from one command surface.
• Portable releases: checksum-verified binaries target macOS, Linux, and Windows, alongside the versioned GitHub Action.

Results

• One entry point for polyglot quality checks while each ecosystem keeps its strongest analyzer.
• Faster feedback through concurrent and changed-file-aware execution.
• Incremental enforcement that lets established repositories block new regressions before paying down every old finding.
• Consistent evidence in pull requests through annotations, summaries, SARIF, and explicit exit behavior.

Why it matters

Quality policy becomes fragile when it lives in tribal knowledge, a collection of unrelated scripts, or a CI file nobody runs locally. It becomes heavy when a wrapper tries to reimplement every analyzer.

quality takes a narrower path: detect repository intent, resolve native tools reproducibly, run them efficiently, and make their results understandable everywhere the work happens.

Read the quality documentation or explore the source on GitHub.