ln-640-pattern-evolution-auditor

What it does

Audits architectural patterns against best practices, maintains a patterns catalog with compliance scores. Use when auditing pattern evolution.

How it works

• The skill orchestrates a multi-phase pattern evolution audit, relying on a core 4-score model (Compliance, Completeness, Quality, Implementation) with a 70% threshold. It coordinates several worker components (ln-641 to ln-647) in Phase 4, with ln-641 scored in Phase 5, and uses mandatory reads from shared references to guide detection, validation, and aggregation.
• Phase 1a Baseline Detection loads and reconciles docs/project/patterns_catalog.md, ensures required sections exist, links ADRs and Guides, and auto-detects patterns in the pattern library as baseline.
• Phase 1b Adaptive Discovery extends the catalog with patterns discovered from structural and document-based heuristics, returning per-pattern evidence and marking them as Discovered Patterns (Adaptive).
• Phase 1c Pattern Recommendations add advisory patterns to the catalog under a dedicated section.
• Phase 1d Applicability Verification checks that detected patterns have sufficient components, plus semantic verification via MCP Ref and WebSearch, and updates catalog sections for Excluded or Verified statuses.
• Phase 2+ involve best-practices research, domain setup, and a multi-worker audit (Phase 4) across potential domains, followed by Phase 5 per-pattern analysis (ln-641).
• Phase 6 performs cross-domain aggregation by parsing worker outputs (642–647), identifying systemic issues, and proposing architectural fixes.
• Phase 7 Gap Analysis enumerates undocumented patterns, missing components, layer violations, consistency issues, and systemic issues.
• Aggregation and reporting rely on templates and core contracts defined in shared references (audit_worker_core_contract.md, audit_worker_report_template.md) and produce per-worker JSON-like reports in designated output paths.

When to use it

Use when auditing pattern evolution, ensuring patterns_catalog.md is up-to-date, and when you need cross-domain and domain-aware insights about pattern compliance, quality, and potential systemic issues.

What it can touch

• Worker orchestration uses agents to invoke these skills: ln-641-pattern-analyzer, ln-642-layer-boundary-auditor, ln-643-api-contract-auditor, ln-644-dep-graph-auditor, ln-645-open-source-replacer, ln-646-project-structure-auditor, ln-647-env-config-auditor. It relies on shared references and templates for evaluation and reporting.

Caveats

• Requires loading mandatory references and templates (e.g., shared/references/task_delegation_pattern.md, two_layer_detection.md, scoring_rules.md, audit_coordinator_domain_mode.md).__n- Outputs are file-based reports with standardized naming per worker (e.g., 641-pattern-{slug}.md, 642-layer-boundary[-{domain}].md, etc.).