Biome

Biome for Frontend Assets in .NET Repositories

Trigger On

• the repo has biome.json, @biomejs/biome, or the user asks for a faster all-in-one frontend formatter-linter stack
• the repo wants one tool for formatting, linting, and import organization across JS, TS, CSS, JSON, GraphQL, or HTML
• the team is comparing Biome against ESLint plus Prettier or wants to simplify the current stack

Do Not Use For

• repos that rely on ESLint plugins or framework-specific rules Biome does not cover yet
• runtime site audits such as headers, accessibility, and SEO; route that to webhint
• cases where a dedicated CSS or HTML tool is still the deliberate owner and no migration is requested

Inputs

• the nearest AGENTS.md
• package.json
• biome.json or biome.jsonc
• current ownership across ESLint, Prettier, Stylelint, and import ordering

Workflow

1. Decide ownership first:
   • Biome as the main formatter and linter
   • Biome only for formatting
   • Biome in coexistence with ESLint for plugin gaps
2. Prefer a repo-local pinned install so CI and developer machines use the same version.
3. Generate biome.json only after confirming what the repo wants Biome to own.
4. Add repeatable scripts to package.json, for example:
   • biome check .
   • biome check . --write
5. Keep file ownership explicit:
   • Biome can own formatting, linting, and import sorting
   • webhint still owns site-runtime audits
   • ESLint may stay for plugin-heavy cases the repo intentionally keeps
6. Start migrations with check and bounded folders before flipping the whole repo to --write.
7. Re-run the frontend build and tests after broad formatting or lint-fix passes.

Current Upstream Notes

• Biome CLI 2.5.12 fixes Astro expression parsing, nested JSX attribute sorting, Unicode stdin/stdout corruption, and repeated dependency scanning. Validate representative Astro/TSX files, Unicode input, and formatter idempotence before dropping workarounds.

• noFloatingPromises now handles cyclic imports and awaited Promise aliases more accurately; awaiting an array of Promises still leaves its elements unhandled. Use Promise.all or an explicit awaited loop when appropriate. New Nursery rules remain opt-in, including useFlatMathMinMax, useModernMathApis, noThisOutsideOfClass, noVueDeprecatedScopedSlots, and noBunModules.

• Biome CLI 2.5.11 includes the 2.5.8-2.5.10 rule and parser additions and adds Nursery rules for unsafe Astro set:html and undeclared CSS custom properties. Enable Nursery rules explicitly; do not treat a version update as permission to change the repository's lint policy.

• The release fixes Astro shorthand anchors, implicit fragments, void elements, and full HTML handling through --stdin-file-path; Vue template/CSS binding false positives; GraphQL naming trivia; and several type-inference performance paths. Re-run the repository's existing biome check and formatter-idempotence tests before removing suppressions or accepting churn.

• When --stdin-file-path is used, verify that the intended nested biome.json is selected and that ignored input produces the expected warning. Keep fixtures for Vue custom blocks, Svelte attachments, Tailwind class sorting, and non-ASCII diagnostic spans when those surfaces matter.

• Recent Biome changes also continue expanding CSS/SCSS, HTML accessibility, import sorting, watch mode, and upgrade-command surfaces; verify actual CLI ownership before replacing ESLint or Stylelint.

Bootstrap When Missing

1. Detect current state:
   • rg --files -g 'package.json' -g 'biome.json*'
   • rg -n '"@biomejs/biome"|"eslint"|"prettier"|"stylelint"' --glob 'package.json' .
2. Prefer a repo-local pinned install:
   • npm i -D -E @biomejs/biome
3. Create config deliberately:
   • npx @biomejs/biome init
4. Add repeatable commands to AGENTS.md and package.json, then verify with:
   • npx @biomejs/biome check .
   • npx @biomejs/biome check . --write
5. Return status: configured if Biome is now wired with explicit ownership, or status: improved if the existing setup was tightened.
6. Return status: not_applicable when the repo intentionally stays on ESLint-centered ownership and no migration or comparison was requested.

Handle Failures

• Missing-rule parity with specialized ESLint plugins is an ownership problem; keep ESLint for those files until the gap is intentionally closed.
• Overly broad --write runs can cause large churn; start with bounded folders or changed files first.
• Generated assets or vendored code should be excluded in biome.json before trusting the signal.
• If developers complain that Biome and ESLint disagree, define file ownership instead of running both broadly on the same surface by accident.

Deliver

• explicit Biome ownership and version pinning
• checked-in config and repeatable check commands
• a migration or coexistence plan versus ESLint and other frontend tools

Validate

• the chosen ownership model is documented
• CI and local runs use the same Biome version
• the target globs exclude generated and vendored assets
• downstream build or test flows still pass after --write runs

Ralph Loop

1. Plan: analyze current state, target outcome, constraints, and risks.
2. Execute one step and produce a concrete delta.
3. Review the result and capture findings.
4. Apply fixes in small batches and rerun checks.
5. Update the plan after each iteration.
6. Repeat until outcomes are acceptable.
7. If a dependency is missing, bootstrap it or return status: not_applicable with a reason.

Required Result Format

• status: complete | clean | improved | configured | not_applicable | blocked
• plan: concise plan and current step
• actions_taken: concrete changes made
• verification: commands, checks, or review evidence
• remaining: unresolved items or none

Example Requests

• "Replace Prettier and basic linting with Biome in this repo."
• "Add Biome to the frontend under ClientApp."
• "Explain whether we should keep ESLint after adding Biome."