feat: refresh shared skills catalog

skills/benchmark-refactor/references/madr-template.md

@@ -0,0 +1,45 @@
+# Markdown Architectural Decision Record (MADR) Template
+
+Save approved architectural and refactoring decisions to `docs/adr/NNNN-[slug].md`.
+
+```markdown
+# ADR-[NNNN]: [Short Title of Architectural Decision]
+
+## Status
+[Proposed | Accepted | Superseded | Deprecated]
+
+## Context and Problem Statement
+[Describe the context, performance bottleneck, or structural limitation prompting this decision.]
+
+## Decision Drivers
+- [Driver 1, e.g. Throughput requirement: sustain >= 25,000 events/sec]
+- [Driver 2, e.g. Latency target: maintain p99 < 15ms]
+- [Driver 3, e.g. Interface stability: zero breaking changes to public API]
+- [Driver 4, e.g. Regression safety: zero test or type check regressions]
+
+## Considered Options
+- **Option 1**: [Short description of candidate A]
+- **Option 2**: [Short description of candidate B]
+- **Option 3**: [Short description of candidate C]
+
+## Decision Outcome
+Chosen Option: [Option X], because [justification based on empirical benchmark results and complexity balance].
+
+## Empirical Benchmark Outcomes
+- **Baseline Execution**:
+  $\mu = XX.X\text{ms} \pm X.X\text{ms}$ (Throughput: $XX,\!XXX\text{ ops/sec}$)
+- **Option 1 Benchmark**:
+  $\mu = XX.X\text{ms} \pm X.X\text{ms}$ ($+XX\%$ improvement; [reason if rejected])
+- **Option 2 Benchmark**:
+  $\mu = XX.X\text{ms} \pm X.X\text{ms}$ ($+XX\%$ improvement, $p < 0.001$; [selected])
+- **Option 3 Benchmark**:
+  [Eliminated due to test failure or infrastructure complexity]
+
+## Positive Consequences
+- [Positive consequence 1]
+- [Positive consequence 2]
+
+## Negative Consequences & Accepted Trade-offs
+- [Negative consequence 1, e.g. Added dependency]
+- [Negative consequence 2, e.g. Increased algorithmic complexity]
+```