Code Audit (Deep)

Line-level audit. Companion to architectural-hotspots. Hotspots ranks files by structural shape (fan-in, fan-out, LOC, cycles); this skill reads files and emits specific, actionable findings with line numbers and fix sketches.

The core failure mode this skill exists to prevent: producing vague "consider refactoring X" advice instead of commit.rs:426 — NonceGen calls SystemTime::now() per nonce; hoist or seed-and-increment. The first is what a graph tool already said; the second is what the user actually wanted.

When to run

Use whenever the user wants findings inside code rather than rankings across files. Triggers include:

• "audit commit.rs" / "review plan.rs"
• "find perf issues in this module"
• "what could go wrong in apply_changes?"
• "deep review of the parser"
• "where is this slow?"
• After running architectural-hotspots and the user wants the next

level of detail on the flagged files.

Do not use when the user wants:

• Architectural overview / rankings → that's architectural-hotspots.
• Style / formatting issues → clippy / eslint / ruff handle these.
• Security review of specific CVE classes → a dedicated security skill

should run. Overlap is fine; this skill won't hunt for known CVEs.

• A single trivial bug fix in a known file — just fix it.

Method

The strength of this skill comes from depth, not breadth. Better to audit three files thoroughly than ten superficially.

1. Pick targets.

• If the user named files, those are the targets.
• If not, and a recent hotspots report exists, take the top 3-5

files by combined signal (god + hub, god + tangle, or any file inside a cycle).

• If neither, ask the user which files — not "should we audit?"

but "which files do you want audited?" so the answer drives work.

1. Read each target end to end. Headers and type signatures are

not enough. A finding requires a line number, and a line number requires having read the line. Skim-reading is the most common failure of this skill — defend against it actively.

1. Audit along the five axes (next section). Most files only hit

two or three axes. Don't reach for findings in an axis that genuinely doesn't apply.

1. Emit findings in the fixed format (later section). Categorized.

file:line. One-line symptom. One-line fix sketch. No paragraphs of prose around each finding — the user is reading for signal.

1. Pick the top 3-4 highest-leverage fixes out of the full list

and name them at the end. This is the deliverable the user will actually act on first.

The five axes

Each axis answers a different question. They are deliberately orthogonal — a finding belongs in exactly one category.

1. Perf

What is making this slow that shouldn't be?

Smells:

• Syscalls in per-element loops. open, stat, exists,

metadata, read_dir called inside a loop body.

• Time / RNG / clock calls per element. SystemTime::now(),

Instant::now(), rand, gettimeofday — fine once, expensive per element.

• Allocation in hot loop. String::new, Vec::new, .clone()

of non-trivial data, .to_owned(), format! in inner iter.

• Full re-scans where incremental would do. Re-running a regex

over the entire post-image to detect changes. Re-walking a tree that was just walked. Recompiling a pattern per call.

• Redundant parse / compile. Same file read twice. Same regex

compiled in two paths. Same tree-sitter Query built twice.

• Forced materialization. .collect::<Vec<_>>() of a stream

whose only consumer is .iter() or one-pass.

Probes (use as starting points, not as the audit itself):

rg -n 'SystemTime::|Instant::|\.now\(\)' <file>
rg -n '\.exists\(\)|\.metadata\(\)|fs::read|read_to_string' <file>
rg -n '\.clone\(\)' <file>
rg -n 'collect::<Vec' <file>

2. Correctness — error handling

Where do errors silently disappear?

Smells:

• let _ = fallible_call(); — return value, including errors,

discarded. Sometimes intentional (best-effort cleanup) but often a hidden bug. If intentional, the comment must explain.

• .ok() immediately after ?-able call — converts to Option,

drops error type.

• First-error-wins aggregation where the caller needs to see

all failures: results.into_iter().find(|r| r.is_err()), .collect::<Result<Vec<_>, _>>() when partial success matters.

• Cleanup / rollback / recover that can itself fail silently.

A rollback function returning () that calls fs::rename internally without surfacing failures is hiding partial-state bugs.

• ? inside a loop where one bad element should not abort the

whole batch (or vice versa — ? missing where it should fail fast).

• match arms that absorb specific err variants without comment.
• fsync / sync_all on a parent dir failing silently — only

observable when the OS later loses data.

Probes:

rg -n 'let _ = ' <file>
rg -n '\.ok\(\);|\.ok\(\)$|\.ok\(\)\?' <file>
rg -n 'unwrap_or\(|unwrap_or_else' <file>
rg -n 'if let Err' <file>      # often where silencing happens

3. Memory shape

What is held in memory that doesn't need to be, or held twice?

Smells:

• Pre- and post- of the same data held together. Struct holds

full new text and the rendered diff of old→new. The diff already encodes the new text relative to old — keeping both doubles per-file footprint.

• Vec<String> where Vec<&str> would suffice given lifetimes.
• Arc<Mutex<T>> where T: Copy + atomics would do.
• **Large enum variants forcing the whole enum to its max

variant.** One 200-byte variant inflates every instance. Box<LargeVariant> fixes.

• Long-lived caches with no eviction policy.
• Whole-file buffers in a tool that could stream.
• Per-file accumulator holding original + new + diff

simultaneously while only one is consumed downstream.

4. Function-level complexity

Where is one function doing too much, or too dangerously?

These are visible only by reading function bodies — graph tools miss them because they live below the file boundary.

Smells:

• Recursion with no depth bound — especially on data the user

controls (parser, walker, AST visitor). Stack-overflow vector.

• Function > ~80 lines doing more than one thing.
• match dispatch with > ~10 arms — table-driven version often

clearer and easier to extend.

• Closure capturing &mut to many outer variables — usually a

struct trying to be born.

• Function with > 5 parameters — parameter object or builder.
• Two functions whose bodies differ only by a constant or branch

— collapse to one.

5. Coupling — semantic, not graph

What couples that the import graph cannot see?

architectural-hotspots sees file-to-file imports. This axis catches the rest:

• Two modules look orthogonal but both call the same set of

helpers from a third — they share an implicit protocol that wants to be made explicit.

• A "library" module that branches on a value it gets from

exactly one caller — the abstraction has one user. Inline it or own the branch in the caller.

• Dual orchestrators with overlapping imports — one

orchestrator can usually absorb the other, or both should delegate to a thinner core. Hotspots flags both as "tangles" without naming the relationship.

• Type defined in module A, used only by module B — wrong home.
• "Generic" helper used only by one site — not generic, just

premature.

Output format

Use this exact template. Counts in the headings let the user scan volume at a glance. Omit any section with zero findings — do not pad.

**Perf (N)**
- `file:line` — symptom in ≤8 words. Fix: <one short phrase>.

**Correctness (N)**
- `file:line` — symptom. Fix: <one short phrase>.

**Memory (N)**
- `file:line` — symptom. Fix: <one short phrase>.

**Complexity (N)**
- `file:line` — symptom. Fix: <one short phrase>.

**Coupling (N)**
- `file_a:line + file_b:line` — symptom. Fix: <one short phrase>.

**Top targets**
3-4 highest-leverage fixes, named with file + symptom.

Examples

Good:

Perf (2) - commit.rs:426 — NonceGen::next calls SystemTime::now() per nonce. Fix: hoist call, increment counter. - commit.rs:328 — recover_sweep does exists() syscall per backup. Fix: single read_dir + filter.

Bad (vague, no line, no fix):

Consider refactoring commit.rs for performance. Some operations may be inefficient.

Bad (over-prosaic, paragraph form):

Looking at commit.rs, around the NonceGen struct, I noticed that it calls SystemTime::now() which is a syscall. This could be slow if called many times, although it depends on the use case…

The format constraint matters because the user reads dozens of findings; signal density wins.

Calibration

A good audit of a single ~500-line file usually yields 5-15 real findings. Fewer than ~3 means either the file is genuinely clean (a real outcome, say so) or the audit was too shallow. More than ~20 usually means the bar dropped — drop the weakest ones rather than pad the list.

Every finding must survive the "would the user act on this?" test:

When in doubt, drop it.

Anti-patterns

• Skim-reading. Defaulting to function signatures and headers,

emitting findings about "what the function probably does". Always read bodies before claiming a finding.

• Overconfident fix sketches. "Fix: use tokio::spawn" when the

codebase is sync. Match fixes to the project's actual stack.

• Findings the codebase already addresses. A // SAFETY: block,

a documented let _ =, a comment explaining why a clone is required — not findings. Read the comments.

• Hand-waving "consider X" verbs. Replace with concrete

imperatives: hoist, inline, collapse, extract, box this variant, bound the recursion, aggregate all errors.

• Trying to be exhaustive. A focused list of 8 real findings

beats a sprawling list of 25 mostly-noise. The user picks 3-4 to act on either way.