Code Quality Auditor (L3 Worker) Skill

Paths: File paths (shared/, references/, ../ln-*) are relative to skills repo root. If not found at CWD, locate this SKILL.md directory and go up one level for repo root. If shared/ is missing, fetch files via WebFetch from https://raw.githubusercontent.com/levnikolaevich/claude-code-skills/master/skills/{path}.

Code Quality Auditor (L3 Worker)

Type: L3 Worker

Specialized worker auditing code complexity, method signatures, algorithms, and constants management.

Purpose & Scope

Audit code quality (Categories 5+6+NEW: Medium Priority)
Check complexity metrics, method signature quality, algorithmic efficiency, constants management
Return structured findings with severity, location, effort, recommendations
Calculate compliance score (X/10) for Code Quality category

Inputs

MANDATORY READ: Load shared/references/audit_worker_core_contract.md.

Receives contextStore with: tech_stack, best_practices, principles, codebase_root, output_dir.

Domain-aware: Supports domain_mode + current_domain (see audit_output_schema.md#domain-aware-worker-output).

Workflow

MANDATORY READ: Load shared/references/two_layer_detection.md for detection methodology.

Parse context -- extract fields, determine scan_path (domain-aware if specified), extract output_dir
Scan codebase for violations (Layer 1)
- All Grep/Glob patterns use scan_path (not codebase_root)
- Graph acceleration (if available): IF contextStore.graph_indexed OR .hex-skills/codegraph/index.db exists:
  - Complexity + God classes: audit_workspace(path=scan_path, verbosity="full") -- use returned hotspots to pre-identify complex functions and god classes.
  - Module metrics: analyze_architecture(path=scan_path, verbosity="full") -- use returned coupling metrics for cascade depth and coupling analysis.
  - Fall back to grep patterns below if graph unavailable.
  - Outline-first read: outline(file_path) before reading large source files -- understand function/class structure for complexity analysis.
- Example: Grep(pattern="if.*if.*if", path=scan_path) for nesting detection
Analyze context per candidate (Layer 2 -- MANDATORY) Layer 1 finding without Layer 2 = NOT a valid finding. Before reporting, ask: "Is this violation intentional or justified by design?"
- Cyclomatic complexity: is complexity from switch/case on enum (valid) or deeply nested conditions (bad)? Enum dispatch -> downgrade to LOW or skip
- O(n^2): read context -- what's n? If bounded (n < 100), downgrade severity
- N+1: read ORM config -- does it have eager loading configured elsewhere? Admin-only endpoint -> downgrade severity
- God class: is it a config/schema/builder class? -> downgrade
- Cascade depth: already traces calls (implicit Layer 2). Orchestrator function -> SEB does NOT apply (see Conflict Resolution in ARCH-AI-SEB)
Collect findings with severity, location, effort, recommendation
- Tag each finding with domain: domain_name (if domain-aware)
Calculate score using penalty algorithm
Write Report: Build full markdown report in memory per shared/templates/audit_worker_report_template.md, write to {output_dir}/ln-624--{domain}.md (or 624-quality.md in global mode) in single Write call
Return Summary: Return minimal summary to coordinator (see Output Format)

Audit Rules (Priority: MEDIUM)

1. Cyclomatic Complexity

What: Too many decision points in single function (> 10)

Detection:

Count if/else, switch/case, ternary, &&, ||, for, while
Use tools: eslint-plugin-complexity, radon (Python), gocyclo (Go)

Severity:

HIGH: Complexity > 20 (extremely hard to test)
MEDIUM: Complexity 11-20 (refactor recommended)
LOW: Complexity 8-10 (acceptable but monitor)
Downgrade when: Enum/switch dispatch, state machines, parser grammars -> downgrade to LOW or skip

Recommendation: Split function, extract helper methods, use early returns

Effort: M-L (depends on complexity)

2. Deep Nesting (> 4 levels)

What: Nested if/for/while blocks too deep

Detection:

Count indentation levels
Pattern: if { if { if { if { if { ... } } } } }

Severity:

HIGH: > 6 levels (unreadable)
MEDIUM: 5-6 levels
LOW: 4 levels
Downgrade when: Nesting from early-return guard clauses (structurally deep but linear logic) -> downgrade

Recommendation: Extract functions, use guard clauses, invert conditions

Effort: M (refactor structure)

3. Long Methods (> 50 lines)

What: Functions too long, doing too much

Detection:

Count lines between function start and end
Exclude comments, blank lines

Severity:

HIGH: > 100 lines
MEDIUM: 51-100 lines
LOW: 40-50 lines (borderline)
Downgrade when: Orchestrator functions with sequential delegation; data transformation pipelines -> downgrade

Recommendation: Split into smaller functions, apply Single Responsibility

Effort: M (extract logic)

4. God Classes/Modules (> 500 lines)

What: Files with too many responsibilities

Detection:

Count lines in file (exclude comments)
Check number of public methods/functions

Severity:

HIGH: > 1000 lines
MEDIUM: 501-1000 lines
LOW: 400-500 lines
Downgrade when: Config/schema/migration files, generated code, barrel/index files -> skip

Recommendation: Split into multiple files, apply separation of concerns

Effort: L (major refactor)

5. Too Many Parameters (> 5)

What: Functions with excessive parameters

Detection:

Count function parameters
Check constructors, methods

Severity:

MEDIUM: 6-8 parameters
LOW: 5 parameters (borderline)
Downgrade when: Builder/options pattern constructor; framework-required signatures (middleware, hooks) -> skip

Recommendation: Use parameter object, builder pattern, default parameters

Effort: S-M (refactor signature + calls)

6. O(n^2) or Worse Algorithms

What: Inefficient nested loops over collections

Detection:

Nested for loops: for (i) { for (j) { ... } }
Nested array methods: arr.map(x => arr.filter(...))

Severity:

HIGH: O(n^2) in hot path (API request handler)
MEDIUM: O(n^2) in occasional operations
LOW: O(n^2) on small datasets (n < 100)
Downgrade when: Bounded n (n < 100 guaranteed by domain); one-time init/migration code -> downgrade to LOW or skip

Recommendation: Use hash maps, optimize with single pass, use better data structures

Effort: M (algorithm redesign)

7. N+1 Query Patterns

What: ORM lazy loading causing N+1 queries

Detection:

Find loops with database queries inside
Check ORM patterns: users.forEach(u => u.getPosts())

Severity:

CRITICAL: N+1 in API endpoint (performance disaster)
HIGH: N+1 in frequent operations
MEDIUM: N+1 in admin panel
Downgrade when: Admin-only endpoint called <=1x/day -> downgrade to LOW. Eager loading configured elsewhere in ORM -> skip

Recommendation: Use eager loading, batch queries, JOIN

Effort: M (change ORM query)

8. Constants Management (NEW)

What: Magic numbers/strings, decentralized constants, duplicates

Detection:

| Issue | Pattern | Example | |-------|---------|---------| | Magic numbers | Hardcoded numbers in conditions/calculations | if (status === 2) | | Magic strings | Hardcoded strings in comparisons | if (role === 'admin') | | Decentralized | Constants scattered across files | MAX_SIZE = 100 in 5 files | | Duplicates | Same value multiple times | STATUS_ACTIVE = 1 in 3 places | | No central file | Missing constants.ts or config.py | No single source of truth |

Severity:

HIGH: Magic numbers in business logic (payment amounts, statuses)
MEDIUM: Duplicate constants (same value defined 3+ times)
MEDIUM: No central constants file
LOW: Magic strings in logging/debugging
Downgrade when: HTTP status codes (200, 404, 500) -> skip. Math constants (0, 1, -1) in algorithms -> skip. Test data -> skip

Recommendation:

Create central constants file (constants.ts, config.py, constants.go)
Extract magic numbers to named constants: const STATUS_ACTIVE = 1
Consolidate duplicates, import from central file
Use enums for related constants

Effort: M (extract constants, update imports, consolidate)

9. Method Signature Quality

What: Poor method contracts reducing readability and maintainability

Detection:

| Issue | Pattern | Example | |-------|---------|---------| | Boolean flag params | >=2 boolean params in signature | def process(data, is_async: bool, skip_validation: bool) | | Too many optional params | >=3 optional params with defaults | def query(db, limit=10, offset=0, sort="id", order="asc") | | Inconsistent verb naming | Different verbs for same operation type in one module | get_user() vs fetch_account() vs load_profile() | | Unclear return type | -> dict, -> Any, -> tuple without TypedDict/NamedTuple | def get_stats() -> dict instead of -> StatsResponse |

Severity:

MEDIUM: Boolean flag params (use enum/strategy), unclear return types
LOW: Too many optional params, inconsistent naming

Recommendation:

Boolean flags: replace with enum, strategy pattern, or separate methods
Optional params: group into config/options dataclass
Naming: standardize verb conventions per module (get_ for sync, fetch_ for async, etc.)
Return types: use TypedDict, NamedTuple, or dataclass instead of raw dict/tuple

Effort: S-M (refactor signatures + callers)

10. Side-Effect Cascade Depth

What: Functions triggering cascading chains of external side-effects (DB writes -> notifications -> metrics -> limits).

Detection: MANDATORY READ: Load shared/references/ai_ready_architecture.md for side-effect markers, false positive exclusions, and opaque sink rules.

Glob **/services/**/*.{py,ts,js,cs,java} to find service files
For each public function: check body for side-effect markers (per reference)
Recursively follow called internal functions for additional markers
Calculate max chain depth from entry point

Severity:

HIGH: cascade_depth >= 4
MEDIUM: cascade_depth = 3
OK: depth <= 2
Downgrade when: Orchestrator/coordinator functions (imports 3+ services AND delegates sequentially) -> skip. Depth from opaque sinks (logging, metrics) -> skip

Conflict Resolution: IF function is an orchestrator/coordinator (imports 3+ services AND delegates to them sequentially) -> ARCH-AI-SEB does NOT apply. Orchestrators are EXPECTED to have multiple side-effect categories. Only flag SEB for leaf functions.

Recommendation: Refactor to flat orchestration -- extract side-effects into independent sink functions. See reference.

Effort: M-L

Output: Also generate summary Pipe/Sink table per module:

| Module | Sinks (0-1) | Shallow Pipes (2) | Deep Pipes (3+) | Sink Ratio | |--------|-------------|-------------------|-----------------|------------|

Scoring Algorithm

MANDATORY READ: Load shared/references/audit_worker_core_contract.md and shared/references/audit_scoring.md.

Output Format

MANDATORY READ: Load shared/references/audit_worker_core_contract.md and shared/templates/audit_worker_report_template.md.

Write JSON summary per shared/references/audit_summary_contract.md. In managed mode the caller passes both runId and summaryArtifactPath; in standalone mode the worker generates its own run-scoped artifact path per shared contract.

Write report to {output_dir}/ln-624--{domain}.md (or 624-quality.md in global mode) with category: "Code Quality" and checks: cyclomatic_complexity, deep_nesting, long_methods, god_classes, too_many_params, quadratic_algorithms, n_plus_one, magic_numbers, method_signatures, cascade_depth.

Return summary per shared/references/audit_summary_contract.md.

When summaryArtifactPath is absent, write the standalone runtime summary under .hex-skills/runtime-artifacts/runs/{run_id}/evaluation-worker/{worker}--{identifier}.json and optionally echo the same summary in structured output.

Report written: .hex-skills/runtime-artifacts/runs/{run_id}/audit-report/ln-624--orders.md
Score: X.X/10 | Issues: N (C:N H:N M:N L:N)

Critical Rules

MANDATORY READ: Load shared/references/audit_worker_core_contract.md.

Do not auto-fix: Report only
Domain-aware scanning: If domain_mode="domain-aware", scan ONLY scan_path (not entire codebase)
Tag findings: Include domain field in each finding when domain-aware
Context-aware: Small functions (n < 100) with O(n^2) may be acceptable
Constants detection: Exclude test files, configs, examples
Metrics tools: Use existing tools when available (ESLint complexity plugin, radon, gocyclo)

Definition of Done

MANDATORY READ: Load shared/references/audit_worker_core_contract.md.

[ ] contextStore parsed (including domain_mode, current_domain, output_dir)
[ ] scan_path determined (domain path or codebase root)
[ ] All 10 checks completed (scoped to scan_path):
- complexity, nesting, length, god classes, parameters, O(n^2), N+1, constants, method signatures, cascade depth
[ ] Findings collected with severity, location, effort, recommendation, domain
[ ] Score calculated
[ ] Report written to {output_dir}/ln-624--{domain}.md (atomic single Write call)
[ ] Summary written per contract

Reference Files

Audit output schema: shared/references/audit_output_schema.md

Version: 3.0.0 Last Updated: 2025-12-23

Agent Skills: Code Quality Auditor (L3 Worker)

Install this agent skill to your local

Skill Files