code-metrics

Measure per-function cyclomatic and cognitive complexity and Halstead difficulty for a change, a path, or the tree, per lane (TypeScript/JavaScript, Python, Bash, Go; C# deferred), from whichever external collector already resolves (lizard, radon, ESLint, sonarjs, gocyclo, gocognit, shellmetrics, multimetric); it installs none. Each number cites its reference: cyclomatic 20 from ISO/IEC 5055:2021 §8.2.117, with 10 (McCabe 1976) and 15 (NIST SP 500-235) selectable; cognitive (Campbell, SonarSource) and Halstead (Halstead 1977) carry none, no standard setting one. A lane with no collector says so and the run continues, an unmeasured value is null not zero, and no finding, severity, or exit-code gate is emitted. Use when: 'how complex is this code', 'cyclomatic complexity', 'cognitive complexity', 'complexity audit', 'is this function too complex', 'Halstead difficulty', 'measure complexity of a change'; for lines per file use /code-metrics:audit-size.

Read coverage artifacts a build already wrote (lcov `.info` including the 2.2 `FNL`/`FNA` records, Cobertura XML, coverage.py JSON, a Go cover profile) and report line coverage per file and function for a change, a path, or the tree, plus CRAP per function (`comp^2 * (1 - cov/100)^3 + comp`, Savoia and Evans 2007) from the sibling complexity numbers. It runs no tests and installs nothing. A missing artifact warns, naming the paths searched, never a silent skip; a function with no executable lines reports null, not 0. Coverage and CRAP default to a null reference (no bar) with its provenance, and no finding, severity, or exit-code gate is emitted. Use when: 'coverage of this change', 'coverage per function', 'CRAP score', 'which functions are complex and untested', 'read the lcov report', 'coverage.xml', 'how covered is this file'; for complexity alone use /code-metrics:audit-complexity; for what these numbers can and cannot say, /code-metrics:principles.

Measure duplicated code as clone classes over the changed files, a path, or the whole tree: each class's duplicated lines and tokens with every instance's file and line range, rolled up per lane (TypeScript/JavaScript, Python, Bash, Go, C#) and per directory, from whichever clone detector already resolves. Replication the target repository declares about itself in a sanctioned-replication registry (a path-within-plugin or a canonical-to-copies cluster line) is subtracted from the total and reported as an exclusion naming the registry line rather than as debt; the report emits no finding, no severity, and no exit-code gate. Use when: 'is this duplicated', 'find copy-paste code', 'clone detection', 'duplication report', 'how much of this change is copied', 'DRY check', 'redundant code', 'duplicated lines in the diff'; for lines per file use /code-metrics:audit-size, and for what a duplication number can and cannot support use /code-metrics:principles.

Measure lines per file for the changed files (the default), explicit paths, or the whole tree (`--all`): total, blank, comment, and code lines when `scc` resolves, total and non-blank lines from a bundled counter otherwise, for every text file, the language lanes (TypeScript/JavaScript, Python, Bash, Go, C#) and everything else (markdown, JSON, YAML, PowerShell) in a catch-all lane. Each file is reported beside a configurable reference with its provenance (default 1000 non-blank lines, the plugin's own number and not ISO-backed; 500 selectable; an ISO/IEC 5055 §8.2.115 function-percentage mode selectable), longest first, and the report never emits a finding, a severity, or an exit-code gate. Use when: 'how long are these files', 'lines per file', 'file size audit', 'which files are too big', 'LOC per file', 'is this file over 1000 lines', 'count lines in the change'; for cyclomatic or cognitive complexity use /code-metrics:audit-complexity.

Measure how much of the code is typed for the changed files, a path, or the whole tree, per file and as a percentage per lane: `type-coverage`'s ratio of identifiers whose type is not `any` for TypeScript, and mypy's `--any-exprs-report` coverage over expressions for Python. Because no standard or CWE anchors this measure, the reference is `null` by design and the number is a trend to watch rather than a bar; the report emits no finding, severity, or exit-code gate. Bash and Go have no comparable collector, and C# is not applicable, because an occurrence count is not comparable to a ratio. A lane whose tool is absent says so with an install hint and the run continues. Use when: 'how much of this is typed', 'type coverage', 'how many anys are in this', 'any usage in the change', 'type debt', 'measure our typing', 'mypy any expressions report'; for cyclomatic or cognitive complexity use /code-metrics:audit-complexity, and for what a measure can and cannot say use /code-metrics:principles.

Metric literacy for what this plugin reports: what cyclomatic complexity, cognitive complexity, Halstead difficulty, lines per file, duplication, coverage, CRAP, and type debt mean, and what none of them can tell a reader. Four reference files carry the definitions and what each collector computes, each bundled reference value with its provenance (why cyclomatic 20, why 22 and 80 have no citation), the CRAP formula with its name history and coverage join, and an annotated bibliography (McCabe, Halstead, NIST SP 500-235, Campbell, Lewis 2013, ISO/IEC 5055 and 25023). Cross-metric caveats live here once; owners of the measures it leaves alone are named behind a presence gate. Use when: 'what does cyclomatic complexity mean', 'is CRAP a real metric', 'why is the cyclomatic reference 20', 'what is a good coverage number', 'which metric should I look at', 'code metrics principles'; for the numbers themselves run the /code-metrics:audit-* skills.

Verify or configure the code-metrics plugin for this repository: `check` probes the interpreter, every configuration layer (user-global, team, local overlay, and the consumer's ecosystem files) for the YAML subset and the tracked-file guard, prints every reference with the layer that supplied it, and probes each collector adapter (a version, or missing with its install hint); `apply` writes the tracked `.claude/code-metrics.yaml` team layer per key, idempotently, never installing a tool and never editing `.gitignore`. Use when: 'set up code-metrics', 'configure code metrics', 'is code-metrics configured', 'which collectors are installed', 'set the cyclomatic reference', 'change the file length reference', 'code-metrics setup', or an audit skill reports a configuration layer it could not read.