levnikolaevich/ln-641-pattern-fitness-auditor

Paths: File paths (references/, ../ln-*) are relative to this skill directory.

Pattern Fitness Auditor

Type: L3 Worker

L3 Worker that analyzes a single architectural pattern against best practices and project needs.

Purpose & Scope

• Analyze ONE pattern per invocation (receives pattern name, locations, best practices)
• Find all implementations in codebase (Glob/Grep)
• Validate implementation exists and works
• Calculate 4 scores: compliance, completeness, quality, implementation
• Identify gaps and issues with severity and effort estimates
• Return structured analysis result
• Emit KEEP_PATTERN, SIMPLIFY_PATTERN, COMPLETE_PATTERN, or REPLACE_PATTERN

Out of Scope:

• Cyclomatic complexity thresholds (>10, >20)
• Method/class length thresholds (>50, >100, >500 lines)
• Quality Score focuses on pattern-specific quality (SOLID within pattern, pattern-level smells), not generic code metrics

Inputs

- pattern: string          # Pattern name (e.g., "Job Processing")
- locations: string[]      # Known file paths/directories
- bestPractices: object    # Best practices from MCP Ref/Context7/WebSearch
- output_dir: string       # e.g., ".hex-skills/runtime-artifacts/runs/{run_id}/audit-report"

Note: If pattern evidence is provided by a managed run, use that evidence instead of rediscovering the same pattern from scratch.

Workflow

Detection policy: use two-layer detection (candidate scan, then context verification); load references/two_layer_detection.md only when the verification method is ambiguous. Tool policy: follow host AGENTS.md MCP preferences; load references/mcp_tool_preferences.md and references/mcp_integration_patterns.md only when host policy is absent or MCP behavior is unclear.

Use hex-graph first when implementation discovery materially improves confidence. Use hex-line first for local code reads when available. If MCP is unavailable, unsupported, or not indexed, continue with built-in Read/Grep/Glob/Bash and state the fallback in the report.

Phase 1: Find Implementations

MANDATORY READ: Load references/pattern_library.md -- use "Pattern Detection (Grep)" table for detection keywords per pattern.

IF pattern.source == "adaptive":
  # Pattern evidence was already provided by the caller
  files = pattern.evidence.files
  SKIP detection keyword search (already done in Phase 1b)
ELSE:
  # Baseline pattern -- use library detection keywords
  files = Glob(locations)
  additional = Grep("{pattern_keywords}", "**/*.{ts,js,py,rb,cs,java}")
  files = deduplicate(files + additional)

Phase 2: Read and Analyze Code

FOR EACH file IN files (limit: 10 key files):
  Read(file)
  Extract: components, patterns, error handling, logging, tests

Phase 3: Calculate 4 Scores

MANDATORY READ: Load references/scoring_rules.md -- follow Detection column for each criterion.

| Score | Source in scoring_rules.md | Max | |-------|---------------------------|-----| | Compliance | "Compliance Score" section -- industry standard, naming, conventions, anti-patterns | 100 | | Completeness | "Completeness Score" section -- required components table (per pattern), error handling, tests | 100 | | Quality | "Quality Score" section -- method length, complexity, code smells, SOLID | 100 | | Implementation | "Implementation Score" section -- compiles, production usage, integration, monitoring | 100 |

Scoring process for each criterion:

1. Run the Detection Grep/Glob from scoring_rules.md
2. If matches found -> add points per criterion
3. If anti-pattern/smell detected -> subtract per deduction table
4. Document evidence: file path + line for each score justification

Phase 4: Identify Issues and Gaps

FOR EACH bestPractice NOT implemented:
  issues.append({
    severity: "HIGH" | "MEDIUM" | "LOW",
    category: "compliance" | "completeness" | "quality" | "implementation",
    issue: description,
    suggestion: how to fix,
    effort: "S" | "M" | "L"
  })

# Layer 2 context check (MANDATORY):
# Deviation documented in code comment or ADR? -> downgrade to LOW
# Pattern intentionally simplified for project scale? -> skip


gaps = {
  missingComponents: required components not found in code,
  inconsistencies: conflicting or incomplete implementations
}

Phase 5: Calculate Score

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

Diagnostic sub-scores (0-100 each) are calculated separately and reported in AUDIT-META for diagnostic purposes only:

• compliance, completeness, quality, implementation

Phase 6: Write Report

MANDATORY READ: Load references/templates/audit_worker_report_template.md.

Write JSON summary per 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.

# Build pattern name slug: "Job Processing" -> "job-processing"
slug = pattern.name.lower().replace(" ", "-")

# Build markdown report in memory with:
# - AUDIT-META (extended: score [penalty-based] + diagnostic score_compliance/completeness/quality/implementation)
# - Checks table (compliance_check, completeness_check, quality_check, implementation_check)
# - Findings table (issues sorted by severity)
# - DATA-EXTENDED: {pattern, codeReferences, gaps, recommendations}

Write to {output_dir}/ln-641--{slug}.md (atomic single Write call)

Phase 7: Return Summary

Report written: .hex-skills/runtime-artifacts/runs/{run_id}/audit-report/ln-641--job-processing.md
Score: 7.9/10 (C:72 K:85 Q:68 I:90) | Issues: 3 (H:1 M:2 L:0)

Critical Rules

Apply the already-loaded references/audit_worker_core_contract.md.

• One pattern only: Analyze only the pattern provided in input
• Read before score: Never score without reading actual code
• Detection-based scoring: Use Grep/Glob patterns from scoring_rules.md, not assumptions
• Effort estimates: Always provide S/M/L for each issue
• Code references: Always include file paths for findings
• Unique angle: Audit one pattern's architectural fitness only. Do not audit generic code quality, dependency topology, package health, or runtime operations.
• Action required: Every finding uses KEEP_PATTERN, SIMPLIFY_PATTERN, COMPLETE_PATTERN, or REPLACE_PATTERN.

Definition of Done

Apply the already-loaded references/audit_worker_core_contract.md.

• [ ] All implementations found via Glob/Grep (using pattern_library.md keywords or adaptive evidence)
• [ ] Key files read and analyzed
• [ ] 4 scores calculated using scoring_rules.md Detection patterns
• [ ] Issues identified with severity, category, suggestion, effort
• [ ] Gaps documented (missing components, inconsistencies)
• [ ] Recommendations provided
• [ ] Report written to {output_dir}/ln-641--{slug}.md (atomic single Write call)
• [ ] Summary written per contract

Reference Files

• Scoring rules: references/scoring_rules.md
• Pattern library: references/pattern_library.md

---

Version: 2.0.0 Last Updated: 2026-02-08