frontend-quality-workflow

Run and fix the frontend static-quality lane — formatting, ESLint, TypeScript, markdownlint, jscpd duplication, and rust-code-analysis metrics — through the profile's make target map, raise-only and w

SKILL.md

GitHub

Frontend Quality Workflow Skill

Profile keys consumed

  • make.format, make.lint, make.lint_eslint, make.lint_tsc, make.lint_md, make.lint_dup, make.lint_metrics, make.lint_deps
  • quality.eslint_errors, quality.eslint_warnings, quality.tsc_errors, quality.markdownlint_errors, quality.jscpd_clones, quality.depcruise_violations, quality.metrics_enforced

Context (Input)

  • Code, doc, or config changes exist in the working directory.
  • You need formatting + the static lint lane green before a commit, review, or PR.
  • Profile loaded from .claude/react-sdlc.yml (run /fe-sdlc-setup if missing).

This is the static-quality lane that the full suite's preflight and lint groups route into. When run under /fe-sdlc-finish-pr, the ci-fixer agent drives this loop to green; invoked directly, follow the same steps yourself. For the whole CI flow (lint plus tests, mutation, and prod-side lanes) use ci-workflow.

Task (Function)

Drive the static-quality lane to a clean exit: run the mutating formatter first, then every read-only lint gate, fixing each failure at the root cause.

Required order — the formatter rewrites files, so it must run alone before any verification, never in parallel with a read-only gate:

make format   # mutating preflight: Prettier + qlty fmt
make lint     # read-only aggregate: eslint, tsc, markdownlint, jscpd, rca-metrics, dependency-cruiser

The target mapped by make.format runs the project's Prettier pass (invoked through the project package manager, framework.package_manager) followed by qlty fmt. The target mapped by make.lint is the aggregate that fans out to the individually mapped sub-targets below.

Success Criteria: the target mapped by make.lint exits 0 after make.format has run. Treat exit status as the contract.

Degrade rule (capability absent): if make.lint is null in the profile (a library/app that ships no aggregate lint target), run the individually mapped sub-targets instead, in order — make.lint_eslint, make.lint_tsc, make.lint_md, make.lint_dup, make.lint_metrics, make.lint_deps — skipping any null entry with an explicit capability-absent note. Common skips across divergent repos:

  • SKIPPED: make.lint_dup when make.lint_dup is null (a repo that ships no jscpd duplication gate).
  • SKIPPED: make.lint_metrics when make.lint_metrics is null (a repo that ships no rust-code-analysis metrics gate).
  • make.lint_eslint may resolve to a different target name per repo (e.g. an ESLint target wired under lint-next); make.format may resolve to a check-only variant (format-check); make.lint_deps may resolve to a dependency-cruiser target under a different name. Always use the mapped target, never the bare tool name.

Qlty bootstrap (tooling absent): if the make.format run fails only because the qlty CLI is missing, install it once, put it on PATH, and re-run — do not skip the formatter or stage repo Qlty config. Prefer a binary preinstalled by the host image; otherwise download the release artifact for the host platform and verify its published checksum before extracting — never fetch-and-execute a remote install script:

command -v qlty >/dev/null || {
  target="x86_64-unknown-linux-gnu" \
    && release="https://github.com/qltysh/qlty/releases/latest/download" \
    && dir="$(mktemp -d)" \
    && curl -fsSL -o "$dir/qlty-$target.tar.xz" "$release/qlty-$target.tar.xz" \
    && curl -fsSL -o "$dir/qlty-$target.tar.xz.sha256" "$release/qlty-$target.tar.xz.sha256" \
    && (cd "$dir" && sha256sum --check "qlty-$target.tar.xz.sha256") \
    && mkdir -p "$HOME/.qlty/bin" \
    && tar -xJf "$dir/qlty-$target.tar.xz" -C "$HOME/.qlty/bin" \
      --strip-components=1 "qlty-$target/qlty"
  rc=$?
  [ -n "${dir:-}" ] && rm -rf "$dir"
  [ "$rc" -eq 0 ] || exit "$rc"   # fail loudly here, not with a later "qlty: not found"
}
export PATH="$HOME/.qlty/bin:$PATH"

Substitute target for the host platform (e.g. aarch64-apple-darwin, whose shasum -a 256 -c replaces sha256sum --check). Do not run qlty init or stage .qlty/qlty.toml unless the task explicitly asks for repository Qlty configuration.

Individual Checks

Each logical gate maps to one profile target. Run the smallest failing check while fixing, then the aggregate before finishing.

Gate Profile target What it enforces
Formatter make.format Prettier + qlty fmt (mutating preflight, runs first)
ESLint make.lint_eslint Lint rules incl. the no-suppression / convention gates
TypeScript make.lint_tsc tsc type-check; zero type errors
Markdown make.lint_md markdownlint over docs and skill/agent frontmatter
Duplication (jscpd) make.lint_dup Copy/paste DRY gate (clone mass over the source root)
Metrics (rca) make.lint_metrics rust-code-analysis complexity/size/Halstead hard-fail policy
Dependencies make.lint_deps dependency-cruiser layer/import-boundary rules

Dependency-cruiser is part of the make.lint aggregate but its findings route to architecture, not this lane.

Execution Steps

Step 1: Format

Run the target mapped by make.format first (always through make — the targets wrap the containerized toolchain and the project-package-manager-managed dependencies; never invoke Prettier, qlty, ESLint, tsc, markdownlint, jscpd, or rust-code-analysis directly on the host). Inspect the files it rewrote.

Step 2: Lint

Run the target mapped by make.lint (or, under the degrade rule, each non-null sub-target in order).

  • Success (exit 0): lane clean → done.
  • Failure: identify the failing gate from the grouped output → Step 3.

Step 3: Fix at the Root Cause

Re-run only the smallest failing check while iterating, and fix the code — never the bar:

Failing gate Re-run via Fix Companion Skill
Formatting drift make.format Let the formatter rewrite; commit the formatted result -
ESLint make.lint_eslint Fix the reported rule at source (no eslint-disable) code-organization
TypeScript make.lint_tsc Make the type contract honest; avoid any (no @ts-ignore) -
Markdown make.lint_md Fix headings, fences, list/line structure; keep skill frontmatter to name + description -
Duplication (jscpd) make.lint_dup Deduplicate — extract a shared style fragment, constant, or factory complexity-management
Metrics (rca) make.lint_metrics Split dense functions/files; extract helpers; replace branch chains with lookup maps complexity-management
Dependencies make.lint_deps Fix the layer/import-boundary violation architecture

Refactoring during fixes: if a failure reveals a structural issue (wrong directory, vague name, a component or hook in the wrong module, hardcoded config), consult code-organization before patching.

Step 4: Re-run

Re-run the failing check, then the aggregate make.lint. Repeat Steps 2-4 until it exits 0. Run make.format again only if a fix touched formatting.

Fix Rules

  • Prefer code changes over disabling rules — fix the root cause, never silence it.
  • Keep TypeScript types honest; avoid any unless an external boundary genuinely requires it.
  • Keep markdown skill/agent frontmatter to name and description only.
  • Split complex components, hooks, and helpers instead of lowering the metrics policy.
  • Re-run the failing check after each focused fix; do not batch unrelated fixes blind.

Line Length Disclosure

Before presenting changes, check changed text files for lines longer than 100 characters. If any exist, tell the user each path:line and the measured character count. Treat this as disclosure, not failure, unless a project gate (markdownlint, ESLint max-len, or the rca size policy) actually fails on it.

Constraints (Parameters)

Thresholds come from quality.* in the profile — NEVER decrease them. This lane's gates are fixed-0 ceilings plus the boolean metrics switch; a profile may never raise a ceiling above 0 nor flip the metrics gate off (raise-only rule). The authoritative policy for the metrics gate is config/metrics-policy.json.

Profile key Shipped default Direction
quality.eslint_errors / quality.eslint_warnings 0 ceiling (fixed)
quality.tsc_errors 0 ceiling (fixed)
quality.markdownlint_errors 0 ceiling (fixed)
quality.jscpd_clones 0 ceiling (fixed)
quality.depcruise_violations 0 ceiling (fixed)
quality.metrics_enforced true bool (must stay true)

DO NOT:

  • Lower a quality.* ceiling or relax eslint.config.mjs, .jscpd.json, config/metrics-policy.json, or the dependency-cruiser config to make violations disappear.
  • Skip a failing gate or commit while a mapped lint target fails.
  • Run the underlying tools outside the mapped make targets (they wrap the containerized toolchain and the project-package-manager-managed dependencies).
  • Add any suppression/ignore directive — eslint-disable, // @ts-ignore, // @ts-nocheck, prettier-ignore, markdownlint-disable, editorconfig-checker-disable, or a jscpd/dependency-cruiser ignore comment — to silence a failure. Fix the code or the type contract instead.

If a rule genuinely cannot apply because of an external constraint, raise it with the user before silencing anything; never silence to land a change.

Format (Output)

Required final state: make.format has run, and the target mapped by make.lint (or, under the degrade rule, every non-null mapped sub-target, skipping capability-absent gates with a note) exits 0.

Verification Checklist

  • make.format ran first (mutating preflight) and its rewrites were inspected.
  • make.lint aggregate (or every non-null sub-target) passed: ESLint, TypeScript, markdown, jscpd, rust-code-analysis metrics, dependency-cruiser.
  • Exit status 0; zero ESLint errors/warnings, zero tsc errors, zero markdown errors, zero jscpd clones, zero dependency-cruiser violations.
  • No quality.* ceiling decreased and quality.metrics_enforced still true.
  • No suppression/ignore directive added anywhere.
  • Any line over 100 characters in a changed file disclosed as path:line + count.

Rollback

If the aggregate's parallel grouping causes issues (interleaved failures, resource contention):

  1. Run the mutating preflight on its own first (make.format), then re-run the read-only make.lint aggregate.
  2. Otherwise run the mapped sub-targets individually in the degrade order: make.lint_eslint, make.lint_tsc, make.lint_md, make.lint_dup, make.lint_metrics, make.lint_deps — skipping capability-absent gates with a note — to isolate the failing one.

Related Skills

  • ci-workflow - Run the full suite (this lane plus tests, mutation, and prod-side lanes)
  • complexity-management - Reduce complexity and resolve duplication when jscpd or rust-code-analysis fails
  • code-organization - Fix structural/naming issues, misplaced components/hooks, or hardcoded config a failure reveals
  • architecture - Resolve dependency-cruiser layer/import-boundary violations
  • quality-standards - Overview of all protected thresholds