Lint Coverage Audit

What it reports

1. Scope

   Counts the files ESLint actually opened against the lintable files on disk, and groups the gap by directory.

   Ignore patterns, a files glob that misses an extension, and a narrow CI invocation all shrink the real count silently. The number in the JSON output is the only honest one.

2. Fatal parse errors

   Lists every file ESLint could not parse.

   A file that fails to parse is a file where no rule ran at all. The usual cause is parserOptions.project pointing at a tsconfig that does not cover the directory.

3. Active rules

   Runs --print-config on a real file from your repo and separates configured rules from active ones.

   A rule set to off is configured, not running. The two counts are usually far apart, and the gap is where upgrades get paid for.

4. Ghost plugins

   Finds plugin prefixes your config references, then probes whether the plugin is installed at all.

   off is the only severity ESLint never resolves, so a config can name plugins that do not exist, indefinitely, without a word. The probe sets one rule to error and reads the failure.

5. Config generator leftovers

   Reads the resolved languageOptions and flags two specific shapes: parserOptions.project set while no type-aware rule is active, and ecmaVersion 5 or sourceType script.

   The first makes you pay for a full TypeScript program on every run that no rule consumes. The second is what migration generators write and nobody re-reads; at ecmaVersion 5 modern globals are absent, which surfaces as no-undef on valid code.

6. Orphan disable comments

   Counts eslint-disable comments and reports how many sit in files ESLint never reads.

   --report-unused-disable-directives only reports on linted files, so it is blind exactly where unnecessary disables collect.

7. CI gate

   Scans your CI definitions for a lint invocation and checks whether its exit code can propagate.

   In sh, a script exits with the status of its last command. A lint call followed by a copy or an upload reports the copy.

What the output looks like

This is a real run against one workspace of the monorepo behind this site, trimmed to fit. The first finding is the one worth internalising: twenty rule entries pointing at a plugin that was never installed, every one of them set to off, so ESLint had never said a word about it.

lint-coverage-audit -- what your ESLint run is not checking
config: eslint.config.ts (flat)

[!] 1 plugin referenced in the config but not installed
    These rules are all set to off, which is the only severity ESLint
    never resolves. The config reads like policy; nothing backs it.
    - @typescript-eslint/* -- 20 rule entries, plugin absent

[!] 2 of 8 eslint-disable comments (25%) sit in files ESLint never reads
    --report-unused-disable-directives only reports on linted files.
    - 2 in src/pages/search.astro

[~] ESLint opened 195 of 220 lintable files (89%)
    25 files were never opened.
    - 15 in src/pages/
    - 9 in src/components/

[ok] 64 active rules of 423 configured
[ok] CI lint invocation looks gated

2 needing attention, 1 worth a look. Nothing was changed.

Eighteen of those twenty entries came from a shared config that switches off formatting rules across plugins, including rules from a plugin this workspace does not have. Nothing in that chain is a bug in any single package. It is just what off does.

Where this came from

Two repos, audited by hand, and the same shapes kept coming back. The write-ups have the full measurements: four ways ESLint reports success while checking nothing covers the CI gate, the scope gap and the off severity, and the flat config migration guide covers the parser-without-plugin trap that started it.

The script exists because those numbers are from private repos and you have no reason to take my word for them. Run it on yours instead.