justasmonkev/lazy-clean/lazy-audit

lazy-review, repo-wide. Scan the whole tree instead of a diff. Rank findings by useful simplification, and do not cut required behavior merely to reduce lines.

Tags

Same as lazy-review:

• delete: dead code, unused flexibility, speculative feature. Replacement: nothing.
• stdlib: hand-rolled thing the standard library ships. Name the function.
• native: dependency or code doing what the platform already does. Name the feature.
• yagni: abstraction with no useful boundary, config nobody sets, layer that exposes no simpler contract.
• shrink: same logic, fewer lines. Show the shorter form.

Hunt

Detect which of TypeScript, JavaScript, Java, Python, Ruby, Rust, and Go the repo actually uses. Read each pinned or installed version from its toolchain file, manifest, lockfile, or runtime and keep replacements compatible. If a needed version fact cannot be checked, say so and do not guess; check the latest release only when the user asks for current-version advice. For TS/JS and Python, TS/JS checks and Python checks name common stdlib replacements and the idioms a simplification must not break.

One caller or one export is not proof of waste. Keep a separate function or file when it names a domain idea, hides tricky logic, isolates a side effect or boundary, earns its keep in tests or readability, or is required by a framework contract. Report it only when none of those hold.

Deps the stdlib or platform already ships, interfaces that isolate no meaningful dependency or change, factories with one product, wrappers that only delegate, dead flags and config, hand-rolled stdlib.

Output

One line per finding, ranked: <tag> <what to cut>. <replacement>. [path]. End with net: -<N> lines, -<M> deps possible. Nothing to cut: Lean already. Ship.

Boundaries

Behavior, edge, and failure tests that catch real regressions are not bloat. Keep meaningful mutation evidence when existing red-green checks do not already prove the risky behavior; mutation evidence is optional when they do. Never flag useful evidence only to make the test tree smaller. Every proposed simplification must preserve behavior; a separate correctness/security audit is not this pass.

Scope: over-engineering and complexity only. Correctness bugs, security holes, and performance are explicitly out of scope. Route them to a normal review pass. Lists findings, applies nothing. One-shot — it sets no mode, so there is nothing to revert. To act on the list, say "apply the findings"; the fixes then follow the lazy ladder and the surgical-changes rule.

One production implementation is not evidence of waste. Before removing an adapter, check what external details or independently changing policy its callers would need to learn. Keep a consumer-owned contract that isolates those details, even when its adapter only forwards. A clean scan or fewer lines does not establish design quality.