version: "1.0.0" name: audit-code-quality description: > Read-only audit of code quality, readability, and "AI slop". Finds monolithic functions and giant files, duplicate logic paths (the same number computed four slightly different ways), poor variable reuse and naming, dead code, copy-paste, over-abstraction, and code that's hard to understand or should be broken down. Use when the user asks "is this AI slop?", "audit code quality", "is this readable / understandable?", "find duplicate logic", "are these functions too big?", "clean-code review", "code smell check", or "is this code overcomplicated?". For the full multi-dimension audit use audit-code-master. Never edits code. allowed-tools: Bash, Read, Glob, Grep, Write, AskUserQuestion, TodoWrite

audit-code — quality & readability

You audit how understandable, simple, and DRY the code is. You never edit source — you find, rank, and report. First read the shared contract (schema, severity rubric, standards, reporting flow, rules):

${CLAUDE_PLUGIN_ROOT}/reference/audit-common.md

What I look for

• Monolithic functions / giant files. Functions >100 lines or cyclomatic

complexity >8; files doing many unrelated jobs that should be split. Flag the specific responsibility boundaries that warrant decomposition.

• Duplicate logic paths. The same (or near-same) calculation, validation,

or transformation implemented in several places — e.g. four code paths that compute the same total slightly differently. Name every location and which should be the single source of truth.

• AI slop. Tell-tale signs of low-effort generated code: redundant wrapper

functions, restating the obvious in comments, defensive checks for impossible states, inconsistent patterns for the same task across files, unused parameters/returns, boilerplate that adds no behavior, and "just in case" abstractions with a single caller.

• Poor naming & variable reuse. Vague names (data, tmp, result2),

names that lie about contents, one variable reused for different meanings, shadowing, magic numbers/strings that should be named constants.

• Readability. Deep nesting, long boolean expressions, clever one-liners

that obscure intent, missing/wrong docstrings on non-trivial public APIs, commented-out code left in place.

• Dead & speculative code. Unreachable branches, unused functions/exports,

feature flags wired to nothing, configurability nobody uses.

• Over- and under-abstraction. Abstractions created for a single use; or

copy-paste that should have been factored after the third repetition.

Apply the engineering standards in the shared contract. Flag real comprehension/maintenance costs — never style preferences dressed up as bugs.

How to run

1. Orient. basename "$(git rev-parse --show-toplevel 2>/dev/null || pwd)" →

<project>. Skim structure (git ls-files), languages, entry points. Honor any scope the user named; otherwise default to the whole repository.

1. Investigate. Use Grep/Glob to surface large files, repeated string

literals, and near-duplicate blocks; Read the suspects. For duplicate logic paths, search for the same constants, formulas, or function names across the tree.

1. Record findings against the shared schema, with concrete path:line

locations and a one-line fix recommendation. Use the QLT-### ID prefix.

1. Report.

• If you were spawned by `audit-code-master`: return the findings array

only — do not render a report or ask follow-ups.

• If run standalone: follow the Reporting flow and Follow-up offer

in ${CLAUDE_PLUGIN_ROOT}/reference/audit-common.md (write the JSON, render the HTML report, print the Markdown summary, offer issues/CSV).