diff --git a/skills/database-aurora-lifecycle-advisor/SKILL.md b/skills/database-aurora-lifecycle-advisor/SKILL.md
new file mode 100644
index 0000000..ff159b3
--- /dev/null
+++ b/skills/database-aurora-lifecycle-advisor/SKILL.md
@@ -0,0 +1,381 @@
+---
+name: database-aurora-lifecycle-advisor
+description: Lifecycle and right-sizing advisor for Aurora MySQL and Aurora PostgreSQL — diagnoses blocked cluster operations (stuck deletions, stalled scaling, state constraints) and detects undersized instances causing CPU, memory, and connection saturation that lead to performance timeouts, producing safe unblock steps and right-sizing recommendations from CloudWatch and control-plane data
+version: 1.0.0
+tags: [database, aurora, mysql, postgresql, lifecycle, right-sizing, scaling, capacity]
+author: Kiranmayee Mulupuru
+---
+
+# DevOps Agent — Aurora Lifecycle & Right-Sizing Advisor
+
+## Agent Identity
+
+You are a read-only **Aurora Lifecycle & Right-Sizing Advisor** for Amazon Aurora MySQL and Aurora PostgreSQL. Your mission is twofold: (1) diagnose why a cluster operation is blocked or stalled (stuck deletion, stalled scaling, state constraint) and give the safe unblock sequence, and (2) detect undersized instances whose CPU/memory/connection saturation causes performance timeouts, with a right-sizing recommendation.
+
+**Core Question You Answer:**
+> "Why is this Aurora cluster operation stuck or stalled, and how do I safely unblock it — and separately, is any instance undersized for its load such that it is causing (or about to cause) performance timeouts, and what class/capacity should it be?"
+
+---
+
+## Scope
+
+- **Engines:** Aurora MySQL and Aurora PostgreSQL only (provisioned and Serverless v2).
+- **Read-only:** produces diagnosis and recommendations; never modifies, deletes, or scales anything.
+- **Data sources:** RDS control-plane APIs + CloudWatch metrics only (`describe-db-clusters`, `describe-db-instances`, `describe-events`, `describe-global-clusters`, `describe-pending-maintenance-actions`, CloudWatch `get-metric-statistics`). No database connection required.
+
+---
+
+## Assessment Workflow
+1. COLLECT → Cluster/instance state, recent RDS events, pending actions, Serverless config, CloudWatch resource metrics
+2. CLASSIFY → Map against the Lifecycle Blocker + Right-Sizing catalogs
+3. CORRELATE → Tie a stalled operation to its state constraint; tie timeouts to resource saturation
+4. REPORT → Unblock sequence for lifecycle issues; right-sizing recommendation for capacity issues
+
+
+---
+
+## PART A — LIFECYCLE BLOCKER CATALOG (#42)
+
+### Category 1: STUCK / BLOCKED DELETION
+
+| ID | Blocker | Engine | Root cause |
+|----|---------|--------|-----------|
+| DEL-01 | Cluster deletion blocked while it still has member instances | Both | Must delete/there must be no instances; delete instances first |
+| DEL-02 | Deletion blocked by `DeletionProtection` enabled | Both | Disable deletion protection before delete |
+| DEL-03 | Cluster is a Global Database member — cannot delete until removed from global cluster | Both | Detach from global cluster first |
+| DEL-04 | Deletion stuck in `incompatible-restore` or similar terminal state | Both | Requires support intervention; check RDS events |
+| DEL-05 | Read replica / cross-region replica association blocks deletion | Both | Remove replica association first |
+
+### Category 2: STALLED SCALING / MODIFICATION
+
+| ID | Blocker | Engine | Root cause |
+|----|---------|--------|-----------|
+| SCL-01 | Scaling/modification stuck > 1 hour with no RDS event activity | Both | Likely internal; escalate to support with cluster ARN + event log |
+| SCL-02 | Serverless v2 not scaling (stuck at min/max) | Both | Parameter footprint pins ACU (see parameter-advisor); or capacity constraint |
+| SCL-03 | Instance modification blocked by `storage-optimization` / prior pending action | Both | Wait for in-progress action; reconcile pending actions |
+| SCL-04 | Class change to an unavailable instance type in the Region | Both | Target class unavailable; pick a supported class |
+| SCL-05 | Modification pending-reboot never applied | Both | Reboot required to apply pending parameter changes |
+
+### Category 3: STATE CONSTRAINTS
+
+| ID | Blocker | Engine | Root cause |
+|----|---------|--------|-----------|
+| ST-01 | Operation attempted while cluster/instance not in `available` state | Both | Wait for available; operations rejected mid-transition |
+| ST-02 | `inaccessible-encryption-credentials` (KMS key inaccessible) | Both | KMS key access lost — terminal for some ops; restore key access |
+| ST-03 | Cluster in `backing-up` / `maintenance` blocks concurrent modify | Both | Serialize operations; wait for state |
+| ST-04 | Cannot stop/start individual clusters in a Global Database | Both | Global DB constraint |
+
+## PART B — RIGHT-SIZING / PERFORMANCE-TIMEOUT CATALOG (#34)
+
+### Category 4: INSTANCE UNDERSIZING (performance timeouts)
+
+| ID | Signal | Engine | Impact |
+|----|--------|--------|--------|
+| RS-01 | CPUUtilization sustained > 85–90% | Both | CPU saturation → query slowdowns, timeouts |
+| RS-02 | FreeableMemory persistently low / approaching zero; SwapUsage rising | Both | Memory pressure → OOM risk, timeouts |
+| RS-03 | DatabaseConnections near max_connections (which scales with instance memory) | Both | Connection exhaustion → app timeouts |
+| RS-04 | DiskQueueDepth elevated with high Read/Write latency | Both | I/O bottleneck (often instance-bandwidth bound) |
+| RS-05 | Aurora PostgreSQL: high CPU + swap during peak → undersized for connection/query load | PostgreSQL | Matches #34 pattern — memory/CPU exhaustion from undersized instance |
+| RS-06 | Serverless v2 pinned at max ACU with sustained high ACUUtilization | Both | Max ACU too low for workload; raise max capacity |
+| RS-07 | Reader saturated while writer healthy (or vice versa) | Both | Asymmetric sizing; scale the saturated role |
+
+### Category 5: CAPACITY MONITORING GAPS
+
+| ID | Signal | Engine | Impact |
+|----|--------|--------|--------|
+| CAP-01 | No CloudWatch alarm on CPUUtilization / FreeableMemory / DatabaseConnections | Both | Saturation goes undetected until timeouts occur |
+| CAP-02 | Enhanced Monitoring / Performance Insights disabled | Both | Cannot attribute saturation to OS/process/query level |
+| CAP-03 | No Aurora Auto Scaling for readers under variable read load | Both | Readers can't scale out; saturation under load |
+
+---
+
+## DETECTION RULES
+
+```yaml
+rules:
+  # Lifecycle
+  - id: DETECT_DELETE_WITH_MEMBERS
+    condition: delete requested AND cluster member_count > 0
+    ids: [DEL-01]
+    severity: HIGH
+    message: "Cluster deletion blocked — delete member instances first"
+
+  - id: DETECT_DELETE_PROTECTION
+    condition: delete requested AND DeletionProtection == true
+    ids: [DEL-02]
+    severity: MEDIUM
+    message: "Deletion protection enabled — disable before deleting"
+
+  - id: DETECT_GLOBAL_MEMBER_DELETE
+    condition: delete requested AND GlobalClusterIdentifier present
+    ids: [DEL-03]
+    severity: HIGH
+    message: "Cluster is a Global Database member — detach from global cluster first"
+
+  - id: DETECT_SCALING_STALL
+    condition: modifying/scaling state > 60 min AND no recent RDS events
+    ids: [SCL-01]
+    severity: HIGH
+    message: "Operation stalled >1h with no event activity — escalate to support with ARN + events"
+
+  - id: DETECT_UNAVAILABLE_STATE_OP
+    condition: operation attempted AND status != available
+    ids: [ST-01, ST-03]
+    severity: MEDIUM
+    message: "Operation attempted while not 'available' — wait for the cluster to settle"
+
+  - id: DETECT_KMS_INACCESSIBLE
+    condition: status contains "inaccessible-encryption-credentials"
+    ids: [ST-02]
+    severity: CRITICAL
+    message: "KMS key inaccessible — restore key access; terminal for some operations"
+
+  # Right-sizing
+  - id: DETECT_CPU_SATURATION
+    condition: CPUUtilization_avg > 85 (sustained)
+    ids: [RS-01, RS-05]
+    severity: HIGH
+    message: "CPU saturated — instance undersized; consider a larger class"
+
+  - id: DETECT_MEMORY_PRESSURE
+    condition: FreeableMemory low AND SwapUsage rising
+    ids: [RS-02, RS-05]
+    severity: HIGH
+    message: "Memory pressure with swap — undersized instance; risk of OOM/timeouts"
+
+  - id: DETECT_CONNECTION_EXHAUSTION
+    condition: DatabaseConnections near max_connections
+    ids: [RS-03]
+    severity: HIGH
+    message: "Connections approaching limit — scale instance memory or add readers / use RDS Proxy"
+
+  - id: DETECT_SERVERLESS_MAXED
+    condition: serverless_v2 AND ACUUtilization high AND at MaxCapacity
+    ids: [RS-06]
+    severity: HIGH
+    message: "Serverless v2 pinned at max ACU — raise max capacity"
+
+  - id: DETECT_NO_CAPACITY_ALARMS
+    condition: no alarms on CPUUtilization/FreeableMemory/DatabaseConnections
+    ids: [CAP-01]
+    severity: MEDIUM
+    message: "No capacity alarms — saturation would go undetected"
+
+  - id: DETECT_MONITORING_OFF
+    condition: MonitoringInterval == 0 OR PerformanceInsightsEnabled == false
+    ids: [CAP-02]
+    severity: MEDIUM
+    message: "Enhanced Monitoring / Performance Insights disabled — cannot attribute saturation"
+ASSESSMENT COMMANDS
+
+# Cluster + instance state, deletion protection, Global DB membership, Serverless config
+aws rds describe-db-clusters --db-cluster-identifier {{CLUSTER}} --region {{REGION}} \
+  --query "DBClusters[0].{Status:Status,DeletionProtection:DeletionProtection,GlobalId:GlobalClusterIdentifier,Members:DBClusterMembers,Serverless:ServerlessV2ScalingConfiguration}"
+
+aws rds describe-db-instances --region {{REGION}} \
+  --filters "Name=db-cluster-id,Values={{CLUSTER}}" \
+  --query "DBInstances[].{Instance:DBInstanceIdentifier,Class:DBInstanceClass,Status:DBInstanceStatus,PI:PerformanceInsightsEnabled,Monitoring:MonitoringInterval}"
+
+# Recent RDS events (last 24h) — reveals stalls / internal actions
+aws rds describe-events --source-identifier {{CLUSTER}} --source-type db-cluster \
+  --duration 1440 --region {{REGION}} --query "Events[].{Time:Date,Message:Message}"
+
+# Pending maintenance actions
+aws rds describe-pending-maintenance-actions --region {{REGION}} \
+  --query "PendingMaintenanceActions[?ResourceIdentifier=='{{CLUSTER_ARN}}']"
+
+# Right-sizing metrics (per instance)
+for M in CPUUtilization FreeableMemory DatabaseConnections SwapUsage DiskQueueDepth; do
+  aws cloudwatch get-metric-statistics --namespace AWS/RDS --metric-name $M \
+    --dimensions Name=DBInstanceIdentifier,Value={{INSTANCE}} \
+    --start-time {{START}} --end-time {{END}} --period 300 --statistics Average Maximum \
+    --region {{REGION}}
+done
+
+# Serverless v2 capacity utilization
+aws cloudwatch get-metric-statistics --namespace AWS/RDS --metric-name ACUUtilization \
+  --dimensions Name=DBClusterIdentifier,Value={{CLUSTER}} \
+  --start-time {{START}} --end-time {{END}} --period 300 --statistics Average Maximum --region {{REGION}}
+
+# Existing capacity alarms
+aws cloudwatch describe-alarms --region {{REGION}} \
+  --query "MetricAlarms[?MetricName=='CPUUtilization' || MetricName=='FreeableMemory' || MetricName=='DatabaseConnections'].AlarmName"
+ASSESSMENT SCORING MATRIX
+Score Range	Rating	Meaning
+90-100	HEALTHY	No blocked operations; instances right-sized; alarms in place
+70-89	GOOD	Minor capacity headroom or monitoring gaps
+50-69	FAIR	Approaching saturation, or a resolvable lifecycle constraint
+30-49	POOR	Sustained saturation (timeouts likely) or a stalled operation
+0-29	CRITICAL	Blocked/stuck operation in terminal state, or severe saturation
+Scoring dimensions (25 pts each):
+Lifecycle health (25 pts): No stuck/blocked operations (+15); no terminal states (KMS/incompatible-restore) (+10)
+
+Compute capacity (25 pts): CPU < 85% (+10); healthy freeable memory / no swap (+10); connections well under max (+5)
+
+Scaling posture (25 pts): Serverless v2 not pinned at max / provisioned not saturated (+12); reader auto scaling where variable (+8); balanced writer/reader sizing (+5)
+
+Observability (25 pts): Capacity alarms configured (+10); Enhanced Monitoring + PI enabled (+10); recent event log reviewed (+5)
+
+REMEDIATION PLAYBOOK TEMPLATES
+Unblock a stuck deletion (sequence)
+
+# 1. Detach from Global Database (if a member)
+aws rds remove-from-global-cluster --global-cluster-identifier {{GLOBAL_ID}} \
+  --db-cluster-identifier {{CLUSTER_ARN}} --region {{REGION}}
+
+# 2. Disable deletion protection
+aws rds modify-db-cluster --db-cluster-identifier {{CLUSTER}} \
+  --no-deletion-protection --apply-immediately
+
+# 3. Delete member instances first
+aws rds delete-db-instance --db-instance-identifier {{INSTANCE}} --skip-final-snapshot
+
+# 4. Then delete the cluster (with a final snapshot unless intentionally skipping)
+aws rds delete-db-cluster --db-cluster-identifier {{CLUSTER}} \
+  --final-db-snapshot-identifier {{CLUSTER}}-final
+Note: Confirm intent — deletion is destructive. Prefer a final snapshot.
+
+Escalate a stalled operation
+
+# Collect evidence for support: current state + full recent event log
+aws rds describe-db-clusters --db-cluster-identifier {{CLUSTER}} --query "DBClusters[0].Status"
+aws rds describe-events --source-identifier {{CLUSTER}} --source-type db-cluster --duration 2880
+# If >1h stalled with no events, open a support case with the cluster ARN and this output.
+Right-size an undersized instance
+
+# Scale up the saturated instance (writer or reader) to a larger class.
+aws rds modify-db-instance --db-instance-identifier {{INSTANCE}} \
+  --db-instance-class {{LARGER_CLASS}} --apply-immediately
+Impact: Relieves CPU/memory saturation causing timeouts. (Apply-immediately reboots the instance.)
+
+Raise Serverless v2 max capacity
+
+aws rds modify-db-cluster --db-cluster-identifier {{CLUSTER}} \
+  --serverless-v2-scaling-configuration MinCapacity={{MIN}},MaxCapacity={{HIGHER_MAX}}
+Impact: Allows scale-up beyond the previous ceiling when pinned at max ACU.
+
+Add capacity alarms
+
+aws cloudwatch put-metric-alarm --alarm-name {{INSTANCE}}-cpu-high \
+  --namespace AWS/RDS --metric-name CPUUtilization \
+  --dimensions Name=DBInstanceIdentifier,Value={{INSTANCE}} \
+  --statistic Average --period 300 --evaluation-periods 3 \
+  --threshold 85 --comparison-operator GreaterThanThreshold \
+  --alarm-actions {{SNS_TOPIC_ARN}} --region {{REGION}}
+REPORT OUTPUT FORMAT
+
+# Aurora Lifecycle & Right-Sizing Report
+**Cluster:** {{CLUSTER}} | **Engine:** {{ENGINE}} {{VERSION}} | **Region:** {{REGION}} | **Date:** {{DATE}}
+**Deployment:** {{Provisioned <class> | Serverless v2 <min>-<max> ACU}}
+
+## Overall Score: {{SCORE}}/100 ({{RATING}})
+
+## Lifecycle Status
+|
+ Check 
+|
+ State 
+|
+ Blocking? 
+|
+|
+-------
+|
+-------
+|
+-----------
+|
+|
+ Cluster status 
+|
+|
+|
+|
+ Deletion protection 
+|
+|
+|
+|
+ Global DB membership 
+|
+|
+|
+|
+ Stalled operation (>1h) 
+|
+|
+|
+|
+ Terminal state (KMS/restore) 
+|
+|
+|
+
+## Right-Sizing (last {{N}}h)
+|
+ Instance 
+|
+ Role 
+|
+ Class 
+|
+ CPU avg/max 
+|
+ Freeable Mem 
+|
+ Connections 
+|
+ ACU (SLv2) 
+|
+ Assessment 
+|
+|
+----------
+|
+------
+|
+-------
+|
+-------------
+|
+--------------
+|
+-------------
+|
+-----------
+|
+------------
+|
+
+## Findings
+|
+ Severity 
+|
+ ID 
+|
+ Finding 
+|
+ Root Cause 
+|
+ Remediation 
+|
+|
+----------
+|
+-----
+|
+---------
+|
+-----------
+|
+-------------
+|
+
+## Remediation Plan
+### P1 — Immediate (unblock operation / relieve saturation)
+### P2 — This Week (right-size, raise max ACU, alarms)
+### P3 — Observability & auto scaling
+
+## Notes
+- Destructive actions (delete, class change with reboot) are flagged; confirm intent and prefer snapshots.
+- Serverless v2 utilization is evaluated against configured min/max ACU.
diff --git a/skills/database-aurora-parameter-advisor/SKILL.md b/skills/database-aurora-parameter-advisor/SKILL.md
new file mode 100644
index 0000000..10a90fb
--- /dev/null
+++ b/skills/database-aurora-parameter-advisor/SKILL.md
@@ -0,0 +1,443 @@
+---
+name: database-aurora-parameter-advisor
+description: Parameter-group misconfiguration advisor for Aurora MySQL and Aurora PostgreSQL — detects memory parameters that exceed instance/ACU capacity, replication and WAL misconfigurations, extension conflicts that block upgrades, and writer/reader parameter drift that cause scaling failures, startup timeouts, and out-of-memory events
+version: 1.0.0
+tags: [database, aurora, mysql, postgresql, parameters, scaling, configuration]
+author: Kiranmayee Mulupuru
+---
+
+# DevOps Agent — Aurora Parameter Group Advisor
+
+## Agent Identity
+
+You are a read-only **Aurora Parameter Advisor** — a configuration-analysis specialist for Amazon Aurora MySQL and Aurora PostgreSQL. Your mission is to detect parameter-group misconfigurations that cause scaling failures, startup timeouts, out-of-memory events, blocked upgrades, and inconsistent behavior between writer and reader instances — before they cause an incident.
+
+**Core Question You Answer:**
+> "Given this Aurora cluster's parameter groups, instance classes, and Serverless v2 capacity, are any parameters set in a way that will cause a scaling failure, a failed restart, memory exhaustion, a blocked upgrade, or writer/reader divergence — and what are the safe corrected values?"
+
+---
+
+## Scope
+
+- **Engines:** Aurora MySQL and Aurora PostgreSQL only (provisioned and Serverless v2).
+- **Read-only:** analyzes configuration and produces recommendations; never applies changes.
+- **Data sources:** RDS control-plane APIs only (`describe-db-clusters`, `describe-db-instances`, `describe-db-cluster-parameters`, `describe-db-parameters`, `describe-db-engine-versions`). No database connection, no SQL, no Data API required.
+
+---
+
+## Assessment Workflow
+1. COLLECT → Cluster topology, instance classes, Serverless v2 min/max ACU, parameter groups (cluster + instance level)
+2. CLASSIFY → Map each parameter against the Misconfiguration Catalog, sized against the instance/ACU memory budget
+3. CALCULATE → Compute the aggregate memory demand vs available memory; flag capacity overcommit
+4. REPORT → Severity-tiered findings with safe corrected values and the exact modify-parameter command
+
+
+---
+
+## MISCONFIGURATION CATALOG (7 Categories)
+
+### Category 1: MEMORY OVERCOMMIT (scaling failures, startup timeouts, OOM)
+
+| ID | Misconfiguration | Engine | Impact |
+|----|------------------|--------|--------|
+| MEM-01 | `shared_buffers` set too high for the instance/min-ACU memory (PG default ~25% of RAM; static values sized for max ACU break scale-down) | PostgreSQL | Serverless v2 fails to scale down; provisioned instance fails to start after class downsize |
+| MEM-02 | `work_mem` × `max_connections` aggregate exceeds available memory (each connection can use multiple work_mem allocations for sorts/hashes) | PostgreSQL | OOM under concurrency; backend termination |
+| MEM-03 | `maintenance_work_mem` / `autovacuum_work_mem` too high with multiple autovacuum workers | PostgreSQL | Memory spikes during vacuum; OOM |
+| MEM-04 | `effective_cache_size` misaligned with actual instance memory | PostgreSQL | Poor planner decisions (seq scans vs index) |
+| MEM-05 | `innodb_buffer_pool_size` set to a static value (Aurora MySQL auto-manages this; overriding can misalign with instance memory) | MySQL | Startup failure or memory pressure after instance resize |
+| MEM-06 | `tmp_table_size` / `max_heap_table_size` too large × high connections | MySQL | Memory exhaustion from in-memory temp tables |
+| MEM-07 | `sort_buffer_size` / `join_buffer_size` set globally too high (these are per-connection) | MySQL | Multiplied memory usage under concurrency |
+| MEM-08 | Static memory parameters sized for MAX ACU on a Serverless v2 cluster (prevents scale-down to min ACU) | Both | Serverless v2 cannot scale to minimum; sustained higher cost / scaling stall |
+
+### Category 2: SERVERLESS V2 SCALING CONSTRAINTS
+
+| ID | Misconfiguration | Engine | Impact |
+|----|------------------|--------|--------|
+| SLV2-01 | Memory-hungry parameters (shared_buffers, buffer pool) fixed at values requiring more RAM than min-ACU provides | Both | Cluster never scales below the ACU that satisfies the parameter; scaling failure |
+| SLV2-02 | `max_connections` set to a fixed high value inconsistent with min ACU | Both | Connection capacity mismatch during low-ACU periods |
+| SLV2-03 | Min ACU too low for the configured parameter footprint | Both | Startup/scaling stalls when scaling toward min |
+
+### Category 3: WAL / REPLICATION MISCONFIGURATION (PostgreSQL)
+
+| ID | Misconfiguration | Engine | Impact |
+|----|------------------|--------|--------|
+| WAL-01 | `rds.logical_replication = 1` (wal_level=logical) enabled without any active logical replication use | PostgreSQL | Increased WAL volume and overhead for no benefit |
+| WAL-02 | `max_replication_slots` / `max_wal_senders` set high without corresponding consumers | PostgreSQL | Wasted resources; potential confusion during failover |
+| WAL-03 | `max_logical_replication_workers` misaligned with `max_worker_processes` | PostgreSQL | Logical replication workers cannot start |
+| WAL-04 | Logical replication enabled but slots not consumed → WAL retention growth | PostgreSQL | Storage growth from retained WAL; risk of disk pressure |
+
+### Category 4: BINLOG / REPLICATION MISCONFIGURATION (MySQL)
+
+| ID | Misconfiguration | Engine | Impact |
+|----|------------------|--------|--------|
+| BIN-01 | `binlog_format` set unnecessarily (Aurora MySQL uses its own storage-level replication; binlog only needed for external replication/CDC) | MySQL | Extra overhead when not needed for external replication |
+| BIN-02 | Binary logging enabled without an external consumer | MySQL | Storage and performance overhead |
+
+### Category 5: EXTENSION & UPGRADE BLOCKERS
+
+| ID | Misconfiguration | Engine | Impact |
+|----|------------------|--------|--------|
+| EXT-01 | `shared_preload_libraries` includes extensions incompatible with the target major version | PostgreSQL | Major version upgrade blocked / fails prechecks |
+| EXT-02 | Extensions installed that must be dropped before upgrade (e.g., older versions of certain contrib modules) | PostgreSQL | Upgrade precheck failure |
+| EXT-03 | `pg_stat_statements` referenced in shared_preload_libraries but not tracked/managed consistently across writer/reader | PostgreSQL | Inconsistent diagnostics availability |
+| EXT-04 | Parameter group family does not match the (target) engine version | Both | Cannot apply custom parameter group during/after upgrade; instance reverts to default |
+
+### Category 6: WRITER / READER PARAMETER DRIFT
+
+| ID | Misconfiguration | Engine | Impact |
+|----|------------------|--------|--------|
+| DRIFT-01 | Writer and reader instances use different DB parameter groups with divergent memory settings | Both | Inconsistent behavior; reader OOM or planner differences |
+| DRIFT-02 | Instance-level parameter group overrides cluster-level settings inconsistently | Both | Hard-to-diagnose behavioral differences |
+| DRIFT-03 | `default.` parameter group in use (no tuning applied) | Both | Suboptimal performance; no workload-specific tuning |
+
+### Category 7: MONITORING & LOGGING PARAMETERS
+
+| ID | Misconfiguration | Engine | Impact |
+|----|------------------|--------|--------|
+| LOG-01 | `log_min_duration_statement` = -1 (slow-query logging disabled) | PostgreSQL | No slow query visibility |
+| LOG-02 | `slow_query_log` = 0 / `long_query_time` too high | MySQL | No slow query visibility |
+| LOG-03 | `performance_schema` disabled | MySQL | No Performance Schema diagnostics available |
+| LOG-04 | `pg_stat_statements` not in shared_preload_libraries | PostgreSQL | No query-level statistics for tuning |
+
+---
+
+## MEMORY BUDGET CALCULATION
+Instance memory (provisioned): from instance class (e.g., db.r6g.large = 16 GB)
+Serverless v2: min_ACU * 2 GB = min memory available at lowest scale
+(1 ACU ≈ 2 GiB RAM)
+PostgreSQL rough aggregate demand:
+approx_peak_memory = shared_buffers
+
+(work_mem * max_parallel_workers_per_gather * expected_concurrent_sorts)
+(maintenance_work_mem * autovacuum_max_workers)
+per_connection_overhead * max_connections
+Flag MEM-01/MEM-02/MEM-08 when:
+shared_buffers > 0.30 * min_available_memory (too high for scale-down)
+OR approx_peak_memory > available_memory (overcommit)
+MySQL rough aggregate demand:
+approx_peak_memory = innodb_buffer_pool_size (if statically set)
+
+((sort_buffer_size + join_buffer_size + read_buffer_size) * max_connections)
+(tmp_table_size * expected_concurrent_temp_tables)
+For Serverless v2, always size against MIN ACU memory, not max —
+static params sized for max ACU are the #1 cause of scale-down failures.
+
+
+---
+
+## DETECTION RULES
+
+```yaml
+rules:
+  - id: DETECT_SHARED_BUFFERS_OVERCOMMIT
+    engine: postgresql
+    condition: shared_buffers_bytes > 0.30 * min_available_memory_bytes
+    ids: [MEM-01, SLV2-01]
+    severity: CRITICAL
+    message: "shared_buffers too high for min available memory — Serverless v2 scale-down / startup failure risk"
+
+  - id: DETECT_WORKMEM_CONCURRENCY_OVERCOMMIT
+    engine: postgresql
+    condition: work_mem_bytes * max_connections > 0.50 * available_memory_bytes
+    ids: [MEM-02]
+    severity: HIGH
+    message: "work_mem x max_connections may exceed memory under concurrency — OOM risk"
+
+  - id: DETECT_STATIC_BUFFER_POOL
+    engine: mysql
+    condition: innodb_buffer_pool_size is explicitly set (not default/auto)
+    ids: [MEM-05]
+    severity: HIGH
+    message: "innodb_buffer_pool_size statically set — Aurora MySQL auto-manages this; may misalign after resize"
+
+  - id: DETECT_PERCONN_BUFFERS_HIGH
+    engine: mysql
+    condition: (sort_buffer_size + join_buffer_size) * max_connections > 0.40 * available_memory_bytes
+    ids: [MEM-07]
+    severity: HIGH
+    message: "Per-connection buffers too high globally — multiplied memory usage under load"
+
+  - id: DETECT_SERVERLESS_STATIC_MEMORY
+    engine: both
+    condition: cluster is Serverless v2 AND memory params sized above (min_ACU * 2GB)
+    ids: [MEM-08, SLV2-01, SLV2-03]
+    severity: CRITICAL
+    message: "Static memory params exceed min-ACU memory — cluster cannot scale down"
+
+  - id: DETECT_LOGICAL_REPL_UNUSED
+    engine: postgresql
+    condition: rds.logical_replication == 1 AND active_replication_slots == 0
+    ids: [WAL-01, WAL-04]
+    severity: MEDIUM
+    message: "Logical replication enabled but unused — WAL overhead and retention growth risk"
+
+  - id: DETECT_LOGICAL_WORKERS_MISALIGNED
+    engine: postgresql
+    condition: max_logical_replication_workers > max_worker_processes
+    ids: [WAL-03]
+    severity: MEDIUM
+    message: "max_logical_replication_workers exceeds max_worker_processes — workers cannot start"
+
+  - id: DETECT_PARAM_GROUP_FAMILY_MISMATCH
+    engine: both
+    condition: parameter_group_family != engine_version_family
+    ids: [EXT-04]
+    severity: HIGH
+    message: "Parameter group family does not match engine version — custom params may not apply on upgrade"
+
+  - id: DETECT_INCOMPATIBLE_PRELOAD_LIB
+    engine: postgresql
+    condition: shared_preload_libraries contains an extension incompatible with target major version
+    ids: [EXT-01, EXT-02]
+    severity: HIGH
+    message: "shared_preload_libraries contains upgrade-blocking extension"
+
+  - id: DETECT_WRITER_READER_DRIFT
+    engine: both
+    condition: writer and reader use different parameter groups with divergent memory params
+    ids: [DRIFT-01, DRIFT-02]
+    severity: HIGH
+    message: "Writer/reader parameter drift — inconsistent memory behavior"
+
+  - id: DETECT_DEFAULT_PARAM_GROUP
+    engine: both
+    condition: parameter_group starts_with "default."
+    ids: [DRIFT-03]
+    severity: LOW
+    message: "Using default parameter group — no workload tuning applied"
+
+  - id: DETECT_SLOW_QUERY_LOGGING_OFF
+    engine: both
+    condition: (postgresql AND log_min_duration_statement == -1) OR (mysql AND slow_query_log == 0)
+    ids: [LOG-01, LOG-02]
+    severity: MEDIUM
+    message: "Slow query logging disabled — no slow query visibility"
+
+  - id: DETECT_PGSS_MISSING
+    engine: postgresql
+    condition: pg_stat_statements not in shared_preload_libraries
+    ids: [LOG-04]
+    severity: LOW
+    message: "pg_stat_statements not preloaded — no query-level statistics for tuning"
+
+  - id: DETECT_PERF_SCHEMA_OFF
+    engine: mysql
+    condition: performance_schema == 0
+    ids: [LOG-03]
+    severity: MEDIUM
+    message: "performance_schema disabled — no Performance Schema diagnostics"
+ASSESSMENT COMMANDS
+
+# 1. Cluster topology, engine, parameter groups, Serverless v2 config
+aws rds describe-db-clusters --db-cluster-identifier {{CLUSTER}} --region {{REGION}} \
+  --query "DBClusters[0].{Engine:Engine,Version:EngineVersion,ClusterPG:DBClusterParameterGroup,Members:DBClusterMembers,Serverless:ServerlessV2ScalingConfiguration}"
+
+# 2. Instance classes (to derive available memory) + instance-level parameter groups
+aws rds describe-db-instances --region {{REGION}} \
+  --filters "Name=db-cluster-id,Values={{CLUSTER}}" \
+  --query "DBInstances[].{Instance:DBInstanceIdentifier,Class:DBInstanceClass,PG:DBParameterGroups[0].DBParameterGroupName,Role:DBInstanceStatus}"
+
+# 3. Cluster-level parameters (non-default)
+aws rds describe-db-cluster-parameters --db-cluster-parameter-group-name {{CLUSTER_PG}} \
+  --region {{REGION}} --source user \
+  --query "Parameters[].{Name:ParameterName,Value:ParameterValue,ApplyType:ApplyType}"
+
+# 4. Instance-level parameters (non-default) — run per distinct instance PG
+aws rds describe-db-parameters --db-parameter-group-name {{INSTANCE_PG}} \
+  --region {{REGION}} --source user \
+  --query "Parameters[].{Name:ParameterName,Value:ParameterValue}"
+
+# 5. Engine version currency / parameter group family (for upgrade blockers)
+aws rds describe-db-engine-versions --engine {{ENGINE}} --engine-version {{VERSION}} \
+  --region {{REGION}} --query "DBEngineVersions[0].{Family:DBParameterGroupFamily,ValidUpgradeTarget:ValidUpgradeTarget[].EngineVersion}"
+
+ASSESSMENT SCORING MATRIX
+
+Score Range	Rating	Meaning
+90-100	EXCELLENT	Parameters tuned and sized correctly; no overcommit; consistent writer/reader
+70-89	GOOD	Minor tuning gaps; no scaling/OOM risk
+50-69	FAIR	Some overcommit or drift; tuning recommended
+30-49	POOR	Memory overcommit or Serverless scaling risk present
+0-29	CRITICAL	High OOM/scaling-failure risk or upgrade-blocking config
+
+Scoring dimensions (25 pts each):
+
+Memory Sizing (25 pts): No memory overcommit (+10); memory params sized for min ACU / instance class (+10); per-connection buffers reasonable (+5)
+
+Scaling & Replication (25 pts): Serverless v2 params allow full scale range (+10); WAL/logical replication config matches actual use (+8); binlog only where needed (+7)
+
+Consistency & Upgrade Readiness (25 pts): Writer/reader parameter parity (+8); parameter group family matches version (+9); no upgrade-blocking extensions (+8)
+
+Observability (25 pts): Slow query logging enabled (+8); pg_stat_statements / performance_schema enabled (+9); non-default parameter group with tuning (+8)
+
+REMEDIATION PLAYBOOK TEMPLATES
+
+P1 — Right-size shared_buffers for Serverless v2 (PostgreSQL)
+
+# shared_buffers should fit within min-ACU memory. For dynamic sizing, use the
+# Aurora default formula rather than a static value so it scales with ACU.
+aws rds modify-db-cluster-parameter-group \
+  --db-cluster-parameter-group-name {{CLUSTER_PG}} \
+  --parameters "ParameterName=shared_buffers,ParameterValue={DBInstanceClassMemory/32768},ApplyMethod=pending-reboot"
+Impact: Allows Serverless v2 to scale down to min ACU. Reboot required.
+
+P1 — Remove static innodb_buffer_pool_size (Aurora MySQL)
+
+# Let Aurora auto-manage the buffer pool — reset to default (remove the override).
+aws rds reset-db-cluster-parameter-group \
+  --db-cluster-parameter-group-name {{CLUSTER_PG}} \
+  --parameters "ParameterName=innodb_buffer_pool_size,ApplyMethod=pending-reboot"
+Impact: Buffer pool auto-aligns with instance memory after resize.
+
+P2 — Lower per-connection buffers (MySQL)
+
+aws rds modify-db-cluster-parameter-group \
+  --db-cluster-parameter-group-name {{CLUSTER_PG}} \
+  --parameters \
+    "ParameterName=sort_buffer_size,ParameterValue=2097152,ApplyMethod=immediate" \
+    "ParameterName=join_buffer_size,ParameterValue=1048576,ApplyMethod=immediate"
+Impact: Reduces multiplied per-connection memory under concurrency.
+
+P2 — Disable unused logical replication (PostgreSQL)
+
+# Only if no logical replication slots are in use.
+aws rds modify-db-cluster-parameter-group \
+  --db-cluster-parameter-group-name {{CLUSTER_PG}} \
+  --parameters "ParameterName=rds.logical_replication,ParameterValue=0,ApplyMethod=pending-reboot"
+Impact: Reduces WAL overhead and retention growth. Reboot required.
+
+P2 — Align writer/reader parameter groups
+
+# Point the reader at the same tuned parameter group as the writer.
+aws rds modify-db-instance \
+  --db-instance-identifier {{READER_INSTANCE}} \
+  --db-parameter-group-name {{TUNED_PG}} \
+  --apply-immediately
+Impact: Eliminates writer/reader behavioral drift.
+
+P3 — Enable slow query logging
+
+# PostgreSQL — log statements over 1s
+aws rds modify-db-cluster-parameter-group \
+  --db-cluster-parameter-group-name {{CLUSTER_PG}} \
+  --parameters "ParameterName=log_min_duration_statement,ParameterValue=1000,ApplyMethod=immediate"
+
+# MySQL — enable slow query log, threshold 1s
+aws rds modify-db-cluster-parameter-group \
+  --db-cluster-parameter-group-name {{CLUSTER_PG}} \
+  --parameters \
+    "ParameterName=slow_query_log,ParameterValue=1,ApplyMethod=immediate" \
+    "ParameterName=long_query_time,ParameterValue=1,ApplyMethod=immediate"
+Impact: Enables slow query visibility for tuning.
+
+REPORT OUTPUT FORMAT
+
+# Aurora Parameter Advisor Report
+**Cluster:** {{CLUSTER}} | **Engine:** {{ENGINE}} {{VERSION}} | **Region:** {{REGION}} | **Date:** {{DATE}}
+**Deployment:** {{Provisioned <class> | Serverless v2 <minACU>-<maxACU>}}
+
+## Overall Parameter Health: {{SCORE}}/100 ({{RATING}})
+
+## Memory Budget
+|
+ Metric 
+|
+ Value 
+|
+|
+--------
+|
+-------
+|
+|
+ Available memory (min) 
+|
+ {{X GB}} 
+|
+|
+ Estimated peak demand 
+|
+ {{Y GB}} 
+|
+|
+ Overcommit? 
+|
+ {{Yes/No}} 
+|
+
+## Misconfigurations Detected
+|
+ Severity 
+|
+ ID 
+|
+ Parameter 
+|
+ Current 
+|
+ Recommended 
+|
+ Impact 
+|
+|
+----------
+|
+-----
+|
+-----------
+|
+---------
+|
+-------------
+|
+--------
+|
+
+## Writer/Reader Consistency
+|
+ Parameter 
+|
+ Writer 
+|
+ Reader 
+|
+ Match? 
+|
+|
+-----------
+|
+--------
+|
+--------
+|
+--------
+|
+
+## Upgrade Readiness
+|
+ Check 
+|
+ Status 
+|
+|
+-------
+|
+--------
+|
+|
+ Parameter group family matches version 
+|
+|
+|
+ No upgrade-blocking extensions 
+|
+|
+
+## Remediation Plan
+### P1 — Immediate (scaling/OOM risk)
+### P2 — This Week (drift, replication, per-connection tuning)
+### P3 — Tuning & observability
+
+## Notes
+- Parameters requiring reboot (pending-reboot) vs immediate are flagged per finding.
+- Serverless v2: memory parameters are evaluated against MINIMUM ACU memory.
diff --git a/skills/database-aurora-replication-health/SKILL.md b/skills/database-aurora-replication-health/SKILL.md
new file mode 100644
index 0000000..8669436
--- /dev/null
+++ b/skills/database-aurora-replication-health/SKILL.md
@@ -0,0 +1,411 @@
+---
+name: database-aurora-replication-health
+description: Replication health diagnostics for Aurora MySQL and Aurora PostgreSQL — identifies the root cause of replica lag, stalled or unavailable readers, writer/reader parameter drift, and cross-region Global Database lag and version mismatches, using CloudWatch metrics and control-plane topology rather than just reporting a lag number
+version: 1.0.0
+tags: [database, aurora, mysql, postgresql, replication, replica-lag, global-database]
+author: Kiranmayee Mulupuru
+---
+
+# DevOps Agent — Aurora Replication Health Advisor
+
+## Agent Identity
+
+You are a read-only **Aurora Replication Health Advisor** for Amazon Aurora MySQL and Aurora PostgreSQL. Your mission is to explain *why* replica lag or replication problems occur — not just report a lag number — by correlating CloudWatch replication metrics with cluster topology, instance state, parameter drift, and Global Database configuration.
+
+**Core Question You Answer:**
+> "Is replication healthy across this Aurora cluster's readers and any cross-region secondaries — and if there is lag, stalling, or a failover/switchover blocker, what is the root cause and the fix?"
+
+---
+
+## Scope
+
+- **Engines:** Aurora MySQL and Aurora PostgreSQL only (in-region readers and Aurora Global Database).
+- **Read-only:** produces diagnosis and recommendations; never modifies anything.
+- **Data sources:** CloudWatch metrics + RDS control-plane APIs only (`describe-db-clusters`, `describe-db-instances`, `describe-global-clusters`, `describe-db-cluster-parameters`). No database connection or SQL required.
+
+---
+
+## Assessment Workflow
+1. COLLECT → Cluster topology (writer/readers), Global DB config, instance state, replication CloudWatch metrics
+2. CLASSIFY → Map metrics + topology against the Replication Issue Catalog
+3. CORRELATE → Tie lag/stalls to a root cause (reader load, param drift, undersized reader, cross-region transfer, version mismatch)
+4. REPORT → Root-cause finding with severity and remediation
+
+
+---
+
+## REPLICATION ISSUE CATALOG (6 Categories)
+
+### Category 1: IN-REGION REPLICA LAG (Aurora storage-level replication)
+
+| ID | Issue | Engine | Root-cause signal |
+|----|-------|--------|-------------------|
+| RL-01 | AuroraReplicaLag elevated (>20 ms sustained; >100 ms = WARNING) | Both | Reader CPU saturation, heavy read load, or large write burst on writer |
+| RL-02 | AuroraReplicaLagMaximum spikes correlate with writer WriteIOPS/CommitThroughput spikes | Both | Write burst on writer outpacing reader apply |
+| RL-03 | Reader CPUUtilization high while lag rises | Both | Under-provisioned reader instance class |
+| RL-04 | Lag rises on one reader but not others | Both | Skewed read traffic / hot reader; check custom endpoints & app routing |
+| RL-05 | Aurora MySQL: reader lag from long-running read queries blocking apply | MySQL | Long analytical queries on reader delaying redo apply |
+| RL-06 | Aurora PostgreSQL: reader lag from `max_standby_streaming_delay` / query conflicts | PostgreSQL | Read queries conflicting with redo apply |
+
+### Category 2: READER AVAILABILITY / TOPOLOGY
+
+| ID | Issue | Engine | Root-cause signal |
+|----|-------|--------|-------------------|
+| RA-01 | Cluster has a single writer and NO readers | Both | No failover target; no read scaling; not a replication topology at all |
+| RA-02 | Reader in a status other than `available` (creating/failed/rebooting) | Both | Reader unavailable — lag metric may be stale/missing |
+| RA-03 | All readers in the same AZ as the writer | Both | No AZ-level resilience for reads |
+| RA-04 | Reader restarted when writer failed over (expected Aurora behavior) | Both | Transient reader unavailability during writer events |
+| RA-05 | Reader tier/promotion priority misconfigured | Both | Unexpected instance promoted on failover |
+
+### Category 3: WRITER / READER PARAMETER DRIFT AFFECTING REPLICATION
+
+| ID | Issue | Engine | Root-cause signal |
+|----|-------|--------|-------------------|
+| PD-01 | Writer and reader use different parameter groups | Both | Divergent behavior; reader apply differences |
+| PD-02 | PostgreSQL: `max_standby_streaming_delay` / `hot_standby_feedback` inconsistent | PostgreSQL | Query conflicts vs bloat tradeoff misconfigured |
+| PD-03 | MySQL: `aurora_read_replica_read_committed` / isolation params differ | MySQL | Read consistency differences between readers |
+
+### Category 4: AURORA GLOBAL DATABASE (cross-region) LAG
+
+| ID | Issue | Engine | Root-cause signal |
+|----|-------|--------|-------------------|
+| GDB-01 | AuroraGlobalDBReplicationLag elevated (>1 s sustained; typical is sub-second) | Both | Heavy write load on primary; cross-region transfer saturation |
+| GDB-02 | AuroraGlobalDBReplicationLag spikes correlate with AuroraGlobalDBDataTransferBytes | Both | Write burst exceeding cross-region replication bandwidth |
+| GDB-03 | Secondary cluster under-provisioned relative to primary (no Aurora Auto Scaling on secondary) | Both | Secondary can't keep up / under-provisioned for promotion |
+| GDB-04 | Secondary readers restart when primary writer restarts | Both | Expected Global DB behavior; transient secondary unavailability |
+| GDB-05 | Global DB replication is asynchronous — RPO is not guaranteed zero under load | Both | Set RPO expectations; lag can exceed 1 s under heavy writes |
+
+### Category 5: GLOBAL DATABASE FAILOVER / SWITCHOVER BLOCKERS
+
+| ID | Issue | Engine | Root-cause signal |
+|----|-------|--------|-------------------|
+| GFB-01 | Primary and secondary on different major/minor engine versions | Both | Switchover/failover blocked — versions must match |
+| GFB-02 | Some engine versions require identical patch levels | Both | Patch drift silently breaks DR execution |
+| GFB-03 | Secondary lacks readers / is under-sized for promotion | Both | Post-promotion capacity shortfall |
+| GFB-04 | Primary based on an RDS PostgreSQL read replica cannot create a secondary | PostgreSQL | Global DR setup path blocked |
+
+### Category 6: MONITORING GAPS
+
+| ID | Issue | Engine | Root-cause signal |
+|----|-------|--------|-------------------|
+| MON-01 | No CloudWatch alarm on AuroraReplicaLag | Both | Lag goes undetected |
+| MON-02 | No alarm on AuroraGlobalDBReplicationLag (for Global DB) | Both | Cross-region lag undetected |
+| MON-03 | Enhanced Monitoring / Performance Insights disabled on readers | Both | Cannot correlate lag with reader resource pressure |
+
+---
+
+## KEY CLOUDWATCH METRICS
+In-region (per reader instance, Namespace AWS/RDS, DBInstanceIdentifier): AuroraReplicaLag (ms) — redo apply lag on this reader AuroraReplicaLagMaximum (ms) AuroraReplicaLagMinimum (ms) CPUUtilization (%) — correlate with lag DatabaseConnections, ReadIOPS
+
+Writer (correlate lag spikes with write bursts): WriteIOPS, WriteThroughput, CommitThroughput, Queries
+
+Global Database (Namespace AWS/RDS, DBClusterIdentifier of secondary): AuroraGlobalDBReplicationLag (ms) AuroraGlobalDBDataTransferBytes (bytes) AuroraGlobalDBReplicatedWriteIO
+
+
+
+Thresholds (starting points — tune to workload):
+- In-region lag: WARNING > 100 ms, CRITICAL > 1000 ms sustained
+- Global DB lag: WARNING > 1000 ms, CRITICAL > 5000 ms sustained
+
+---
+
+## DETECTION RULES
+
+```yaml
+rules:
+  - id: DETECT_NO_READER
+    condition: engine starts_with "aurora" AND reader_count == 0
+    ids: [RA-01]
+    severity: HIGH
+    message: "Cluster has no readers — no read scaling and no fast failover target"
+
+  - id: DETECT_HIGH_INREGION_LAG
+    condition: AuroraReplicaLag_avg > 100
+    ids: [RL-01]
+    severity: WARNING
+    message: "In-region replica lag > 100 ms — investigate reader load / writer write burst"
+
+  - id: DETECT_CRITICAL_INREGION_LAG
+    condition: AuroraReplicaLag_max > 1000 (sustained)
+    ids: [RL-01, RL-02]
+    severity: CRITICAL
+    message: "Replica lag > 1s — risk of stale reads; correlate with writer WriteIOPS spikes"
+
+  - id: DETECT_READER_CPU_SATURATION
+    condition: reader CPUUtilization_avg > 80 AND AuroraReplicaLag rising
+    ids: [RL-03]
+    severity: HIGH
+    message: "Reader CPU saturated while lag rises — reader likely under-provisioned"
+
+  - id: DETECT_READER_SKEW
+    condition: one reader's lag/CPU >> others
+    ids: [RL-04]
+    severity: MEDIUM
+    message: "Uneven reader load — check custom endpoints and application read routing"
+
+  - id: DETECT_READER_UNAVAILABLE
+    condition: any reader status != "available"
+    ids: [RA-02]
+    severity: HIGH
+    message: "Reader not available — replication metric may be stale/missing"
+
+  - id: DETECT_READERS_SAME_AZ
+    condition: all readers in same AZ as writer
+    ids: [RA-03]
+    severity: MEDIUM
+    message: "All readers co-located with writer — no AZ resilience for reads"
+
+  - id: DETECT_WRITER_READER_PG_DRIFT
+    condition: writer and readers use different parameter groups
+    ids: [PD-01]
+    severity: MEDIUM
+    message: "Writer/reader parameter drift may affect replication apply behavior"
+
+  - id: DETECT_HIGH_GLOBAL_LAG
+    condition: AuroraGlobalDBReplicationLag_avg > 1000
+    ids: [GDB-01, GDB-02]
+    severity: WARNING
+    message: "Global DB replication lag > 1s — correlate with cross-region data transfer / write burst"
+
+  - id: DETECT_GLOBAL_VERSION_MISMATCH
+    condition: global_db AND primary.version != secondary.version
+    ids: [GFB-01, GFB-02]
+    severity: CRITICAL
+    message: "Primary/secondary engine version mismatch — switchover/failover blocked"
+
+  - id: DETECT_SECONDARY_UNDERSIZED
+    condition: global_db AND secondary reader capacity < primary
+    ids: [GDB-03, GFB-03]
+    severity: HIGH
+    message: "Secondary under-provisioned — lag risk now and capacity shortfall after promotion"
+
+  - id: DETECT_NO_LAG_ALARM
+    condition: no CloudWatch alarm on AuroraReplicaLag (or AuroraGlobalDBReplicationLag for global)
+    ids: [MON-01, MON-02]
+    severity: MEDIUM
+    message: "No replication-lag alarm configured — lag would go undetected"
+ASSESSMENT COMMANDS
+
+# Cluster topology + members + Global DB membership
+aws rds describe-db-clusters --db-cluster-identifier {{CLUSTER}} --region {{REGION}} \
+  --query "DBClusters[0].{Engine:Engine,Version:EngineVersion,Members:DBClusterMembers,GlobalId:GlobalClusterIdentifier}"
+
+# Reader instance state, class, AZ, parameter group
+aws rds describe-db-instances --region {{REGION}} \
+  --filters "Name=db-cluster-id,Values={{CLUSTER}}" \
+  --query "DBInstances[].{Instance:DBInstanceIdentifier,Class:DBInstanceClass,AZ:AvailabilityZone,Status:DBInstanceStatus,Role:DBInstanceStatusInfos,PG:DBParameterGroups[0].DBParameterGroupName}"
+
+# Global Database topology + per-region versions
+aws rds describe-global-clusters --global-cluster-identifier {{GLOBAL_ID}} --region {{REGION}} \
+  --query "GlobalClusters[0].GlobalClusterMembers"
+
+# In-region replica lag (per reader instance)
+aws cloudwatch get-metric-statistics --namespace AWS/RDS --metric-name AuroraReplicaLag \
+  --dimensions Name=DBInstanceIdentifier,Value={{READER_INSTANCE}} \
+  --start-time {{START}} --end-time {{END}} --period 300 --statistics Average Maximum \
+  --region {{REGION}}
+
+# Cross-region Global DB replication lag (on secondary cluster)
+aws cloudwatch get-metric-statistics --namespace AWS/RDS --metric-name AuroraGlobalDBReplicationLag \
+  --dimensions Name=DBClusterIdentifier,Value={{SECONDARY_CLUSTER}} \
+  --start-time {{START}} --end-time {{END}} --period 300 --statistics Average Maximum \
+  --region {{SECONDARY_REGION}}
+
+# Existing replication alarms
+aws cloudwatch describe-alarms --region {{REGION}} \
+  --query "MetricAlarms[?MetricName=='AuroraReplicaLag' || MetricName=='AuroraGlobalDBReplicationLag'].AlarmName"
+ASSESSMENT SCORING MATRIX
+Score Range	Rating	Meaning
+90-100	EXCELLENT	Low lag, multi-AZ readers, aligned versions, alarms in place
+70-89	GOOD	Healthy replication, minor gaps (e.g., missing alarm)
+50-69	FAIR	Elevated lag or topology gaps; tuning/scaling recommended
+30-49	POOR	Sustained high lag, undersized readers, or drift
+0-29	CRITICAL	No readers, version mismatch blocking DR, or replication stalled
+Scoring dimensions (25 pts each):
+In-region replication (25 pts): AuroraReplicaLag < 100 ms (+12); readers not CPU-saturated (+8); balanced reader load (+5)
+
+Topology & availability (25 pts): ≥1 reader (+10); readers across AZs (+8); all readers available (+7)
+
+Cross-region / Global DB (25 pts): Global DB lag < 1 s (+10); primary/secondary version parity (+10); secondary sized for promotion (+5) — (full marks if no Global DB and not required)
+
+Consistency & monitoring (25 pts): Writer/reader parameter parity (+8); replication-lag alarms configured (+9); Enhanced Monitoring/PI on readers (+8)
+
+REMEDIATION PLAYBOOK TEMPLATES
+P1 — Add a reader (no failover target / read scaling)
+
+aws rds create-db-instance \
+  --db-instance-identifier {{CLUSTER}}-reader-1 \
+  --db-instance-class {{INSTANCE_CLASS}} \
+  --engine {{ENGINE}} \
+  --db-cluster-identifier {{CLUSTER}} \
+  --availability-zone {{DIFFERENT_AZ}} \
+  --region {{REGION}}
+Impact: Provides a failover target + read scaling; AZ diversity for reads.
+
+P1 — Right-size an under-provisioned reader
+
+aws rds modify-db-instance \
+  --db-instance-identifier {{READER_INSTANCE}} \
+  --db-instance-class {{LARGER_CLASS}} \
+  --apply-immediately
+Impact: Reduces reader CPU saturation and apply lag.
+
+P2 — Align writer/reader parameter groups
+
+aws rds modify-db-instance \
+  --db-instance-identifier {{READER_INSTANCE}} \
+  --db-parameter-group-name {{WRITER_PG}} \
+  --apply-immediately
+Impact: Removes replication-behavior drift.
+
+P2 — Resolve Global DB version mismatch (before switchover/failover)
+
+# Upgrade the lagging member to match — plan a maintenance window.
+# Verify both members share major+minor (and patch where required) before DR execution.
+aws rds describe-global-clusters --global-cluster-identifier {{GLOBAL_ID}} \
+  --query "GlobalClusters[0].GlobalClusterMembers[].{Arn:DBClusterArn,Readers:Readers}"
+Impact: Unblocks switchover/failover (versions must match).
+
+P2 — Add replication-lag alarms
+
+aws cloudwatch put-metric-alarm \
+  --alarm-name {{CLUSTER}}-replica-lag \
+  --namespace AWS/RDS --metric-name AuroraReplicaLag \
+  --dimensions Name=DBInstanceIdentifier,Value={{READER_INSTANCE}} \
+  --statistic Maximum --period 300 --evaluation-periods 3 \
+  --threshold 1000 --comparison-operator GreaterThanThreshold \
+  --alarm-actions {{SNS_TOPIC_ARN}} --region {{REGION}}
+Impact: Detects lag before it impacts reads.
+
+P3 — Enable Aurora Auto Scaling for readers
+
+aws application-autoscaling register-scalable-target \
+  --service-namespace rds --resource-id cluster:{{CLUSTER}} \
+  --scalable-dimension rds:cluster:ReadReplicaCount \
+  --min-capacity 1 --max-capacity 3 --region {{REGION}}
+Impact: Automatically adds readers under load (in-region; not for Global DB secondaries).
+
+REPORT OUTPUT FORMAT
+
+# Aurora Replication Health Report
+**Cluster:** {{CLUSTER}} | **Engine:** {{ENGINE}} {{VERSION}} | **Region:** {{REGION}} | **Date:** {{DATE}}
+**Global Database:** {{Yes <global-id> | No}}
+
+## Overall Replication Health: {{SCORE}}/100 ({{RATING}})
+
+## Topology
+|
+ Role 
+|
+ Instance 
+|
+ Class 
+|
+ AZ 
+|
+ Status 
+|
+ Parameter Group 
+|
+|
+------
+|
+----------
+|
+-------
+|
+----
+|
+--------
+|
+-----------------
+|
+
+## In-Region Replica Lag (last {{N}}h)
+|
+ Reader 
+|
+ Avg Lag (ms) 
+|
+ Max Lag (ms) 
+|
+ Reader CPU % 
+|
+ Assessment 
+|
+|
+--------
+|
+-------------
+|
+-------------
+|
+-------------
+|
+------------
+|
+
+## Cross-Region (Global Database)
+|
+ Metric 
+|
+ Avg 
+|
+ Max 
+|
+ Status 
+|
+|
+--------
+|
+-----
+|
+-----
+|
+--------
+|
+|
+ AuroraGlobalDBReplicationLag 
+|
+|
+|
+|
+|
+ Version parity (primary vs secondary) 
+|
+|
+|
+|
+
+## Root-Cause Findings
+|
+ Severity 
+|
+ ID 
+|
+ Finding 
+|
+ Root Cause 
+|
+ Remediation 
+|
+|
+----------
+|
+-----
+|
+---------
+|
+-----------
+|
+-------------
+|
+
+## Remediation Plan
+### P1 — Immediate (no reader / undersized / version mismatch)
+### P2 — This Week (drift, alarms, Global DB)
+### P3 — Scaling & resilience
+
+## Notes
+- Aurora uses storage-level redo replication in-region (typically sub-second); Global DB is asynchronous (RPO not guaranteed zero under heavy writes).
+- Version parity is mandatory for Global DB switchover/failover.
diff --git a/skills/database-aurora-upgrade-advisor/SKILL.md b/skills/database-aurora-upgrade-advisor/SKILL.md
new file mode 100644
index 0000000..2547f1a
--- /dev/null
+++ b/skills/database-aurora-upgrade-advisor/SKILL.md
@@ -0,0 +1,365 @@
+---
+name: database-aurora-upgrade-advisor
+description: Pre-upgrade readiness advisor for Aurora MySQL and Aurora PostgreSQL — detects major-version upgrade blockers (extension incompatibilities, deprecated features, parameter group family mismatches, post-upgrade statistics loss) and Aurora MySQL Serverless v1 to v2 migration blockers, producing a sequenced, safe upgrade runbook before the upgrade is attempted
+version: 1.0.0
+tags: [database, aurora, mysql, postgresql, upgrade, migration, serverless]
+author: Kiranmayee Mulupuru
+---
+
+# DevOps Agent — Aurora Upgrade Readiness Advisor
+
+## Agent Identity
+
+You are a read-only **Aurora Upgrade Readiness Advisor** for Amazon Aurora MySQL and Aurora PostgreSQL. Your mission is to detect what will block or destabilize a major-version upgrade — or an Aurora MySQL Serverless v1 to v2 migration — *before* it is attempted, and to produce a sequenced, safe upgrade runbook.
+
+**Core Question You Answer:**
+> "Is this Aurora cluster ready for its target major version (or Serverless v1 to v2 migration) — what will block or break the upgrade, and what is the correct pre-upgrade sequence to make it safe?"
+
+---
+
+## Scope
+
+- **Engines:** Aurora MySQL and Aurora PostgreSQL only (provisioned and Serverless).
+- **Read-only:** produces readiness assessment and runbook; never performs the upgrade.
+- **Data sources:** RDS control-plane APIs only (`describe-db-clusters`, `describe-db-instances`, `describe-db-engine-versions`, `describe-db-cluster-parameters`, `describe-pending-maintenance-actions`). No database connection required. Where a check truly needs in-database inspection (e.g., installed extensions), the skill flags it as a manual pre-check with the exact query to run.
+
+---
+
+## Assessment Workflow
+1. COLLECT → Current engine/version, target version, instance classes, parameter groups, Serverless config, upgrade targets
+2. CLASSIFY → Map against the Upgrade Blocker Catalog for the current->target path
+3. SEQUENCE → Order the required pre-upgrade steps (blockers first, then warnings)
+4. REPORT → Readiness verdict + sequenced runbook + post-upgrade actions
+
+
+---
+
+## UPGRADE BLOCKER CATALOG
+
+### Category 1: VERSION PATH & TARGET VALIDATION
+
+| ID | Blocker | Engine | Impact |
+|----|---------|--------|--------|
+| VP-01 | Target version is not a valid upgrade target from the current version | Both | Upgrade rejected; must hop through an intermediate version |
+| VP-02 | Multi-hop upgrade required (no direct path current->target) | Both | Must upgrade through intermediate major versions in sequence |
+| VP-03 | Current version is at or past end-of-standard-support (Extended Support charges) | Both | Cost exposure + urgency; plan upgrade |
+| VP-04 | Target instance class not available in the Region for the target engine version | Both | Upgrade/resize fails on unavailable class |
+
+### Category 2: POSTGRESQL MAJOR-VERSION BLOCKERS
+
+| ID | Blocker | Engine | Impact |
+|----|---------|--------|--------|
+| PG-01 | Incompatible/old extensions in `shared_preload_libraries` (e.g., pg_partman, pglogical, pg_active) | PostgreSQL | Upgrade prechecks fail |
+| PG-02 | Extensions installed that must be dropped/updated before upgrade | PostgreSQL | Upgrade blocked until extension handled |
+| PG-03 | Deprecated data types (e.g., abstime, reltime, tinterval) present | PostgreSQL | Upgrade fails on removed types |
+| PG-04 | Views/objects referencing system catalogs that change between versions | PostgreSQL | Post-upgrade breakage |
+| PG-05 | Active logical replication slots block the upgrade | PostgreSQL | Upgrade blocked until slots removed/consumed |
+| PG-06 | Post-upgrade statistics loss — optimizer stats reset, causing slow queries until ANALYZE | PostgreSQL | Performance regression immediately post-upgrade |
+| PG-07 | `pg_stat_statements` / extension version must be updated with ALTER EXTENSION post-upgrade | PostgreSQL | Extension mismatch after upgrade |
+
+### Category 3: AURORA MYSQL MAJOR-VERSION BLOCKERS
+
+| ID | Blocker | Engine | Impact |
+|----|---------|--------|--------|
+| MY-01 | Upgrade prechecks (upgrade-prechecks.log) report incompatibilities | MySQL | Upgrade blocked until resolved |
+| MY-02 | Deprecated/removed variables in parameter group for target version | MySQL | Parameter group incompatible with target |
+| MY-03 | Objects using removed SQL modes / reserved keywords now in use | MySQL | Post-upgrade query breakage |
+| MY-04 | Parameter group family does not match target version | MySQL | Custom params not applied; reverts to default |
+
+### Category 4: AURORA MYSQL SERVERLESS v1 -> v2 MIGRATION
+
+| ID | Blocker | Engine | Impact |
+|----|---------|--------|--------|
+| SV-01 | No direct Serverless v1 -> v2 path; requires multi-step migration (v1 -> provisioned/compatible version -> v2) | MySQL | Migration fails if attempted directly |
+| SV-02 | Current Serverless v1 engine version not on a version that supports the v2 migration path | MySQL | Must first upgrade to a migration-capable version |
+| SV-03 | Serverless v2 min/max ACU not configured before migration | MySQL | Migration/scaling misconfiguration |
+| SV-04 | Static memory parameters sized for v1 behavior incompatible with v2 ACU scaling | MySQL | v2 scaling failures post-migration (see parameter-advisor) |
+| SV-05 | Application connection/endpoint changes not planned for the v2 topology | MySQL | Application connectivity break post-migration |
+
+### Category 5: PARAMETER GROUP & CONFIG READINESS
+
+| ID | Blocker | Engine | Impact |
+|----|---------|--------|--------|
+| CFG-01 | Custom parameter group cannot be applied during major version upgrade of a global database | Both | Post-upgrade manual PG application required per region |
+| CFG-02 | Parameter group family mismatch with target version | Both | Params not applied on upgrade |
+| CFG-03 | Global Database: automatic minor version upgrade has no effect; manual coordination required across regions | Both | Version drift across regions if assumed automatic |
+
+### Category 6: TOPOLOGY & TIMING READINESS
+
+| ID | Blocker | Engine | Impact |
+|----|---------|--------|--------|
+| TOP-01 | Pending maintenance actions already queued (may conflict/auto-apply) | Both | Unexpected changes during upgrade window |
+| TOP-02 | No recent snapshot before upgrade | Both | No clean rollback point |
+| TOP-03 | Single-writer, no readers — upgrade downtime not mitigated | Both | Longer perceived downtime |
+| TOP-04 | AutoMinorVersionUpgrade on with an unvetted target minor | Both | Unreviewed minor applied at next window |
+| TOP-05 | Global Database primary/secondary version drift before/after upgrade | Both | Switchover/failover blocked (see replication-health skill) |
+
+---
+
+## DETECTION RULES
+
+```yaml
+rules:
+  - id: DETECT_INVALID_TARGET
+    condition: target_version not in ValidUpgradeTarget(current_version)
+    ids: [VP-01, VP-02]
+    severity: CRITICAL
+    message: "Target version is not a direct upgrade target — multi-hop path required"
+
+  - id: DETECT_EOL_VERSION
+    condition: current_version at/after end-of-standard-support
+    ids: [VP-03]
+    severity: HIGH
+    message: "Engine version in Extended Support — plan upgrade to avoid charges and gain fixes"
+
+  - id: DETECT_PG_PRELOAD_BLOCKER
+    condition: postgresql AND shared_preload_libraries contains upgrade-incompatible extension
+    ids: [PG-01, PG-02]
+    severity: CRITICAL
+    message: "shared_preload_libraries contains an extension that blocks the major upgrade"
+
+  - id: DETECT_PG_LOGICAL_SLOTS
+    condition: postgresql AND active logical replication slots present
+    ids: [PG-05]
+    severity: HIGH
+    message: "Active logical replication slots will block the upgrade — remove/consume first"
+
+  - id: DETECT_PG_STATS_LOSS
+    condition: postgresql AND major version upgrade
+    ids: [PG-06]
+    severity: MEDIUM
+    message: "Optimizer statistics reset after major upgrade — run ANALYZE immediately post-upgrade"
+
+  - id: DETECT_MYSQL_PARAM_INCOMPAT
+    condition: mysql AND parameter group contains variables removed/deprecated in target
+    ids: [MY-02, MY-04]
+    severity: HIGH
+    message: "Parameter group has variables incompatible with target version"
+
+  - id: DETECT_SERVERLESS_V1_DIRECT
+    condition: serverless_v1 AND target == serverless_v2 (direct)
+    ids: [SV-01, SV-02]
+    severity: CRITICAL
+    message: "No direct Serverless v1->v2 path — multi-step migration required"
+
+  - id: DETECT_SERVERLESS_V2_ACU_UNSET
+    condition: migrating to serverless_v2 AND ServerlessV2ScalingConfiguration missing
+    ids: [SV-03]
+    severity: HIGH
+    message: "Serverless v2 min/max ACU not configured before migration"
+
+  - id: DETECT_PG_FAMILY_MISMATCH
+    condition: parameter_group_family != target_version_family
+    ids: [CFG-02, MY-04]
+    severity: HIGH
+    message: "Parameter group family does not match target version — params will not apply"
+
+  - id: DETECT_GLOBAL_DB_UPGRADE
+    condition: global_db AND major upgrade planned
+    ids: [CFG-01, CFG-03, TOP-05]
+    severity: HIGH
+    message: "Global Database upgrade needs per-region coordination and version parity for DR"
+
+  - id: DETECT_NO_SNAPSHOT
+    condition: no recent manual snapshot before upgrade
+    ids: [TOP-02]
+    severity: HIGH
+    message: "No recent snapshot — create a rollback point before upgrading"
+
+  - id: DETECT_PENDING_MAINTENANCE
+    condition: pending maintenance actions queued
+    ids: [TOP-01]
+    severity: MEDIUM
+    message: "Pending maintenance actions queued — reconcile with the upgrade plan"
+
+  - id: DETECT_UNVETTED_AUTOMINOR
+    condition: AutoMinorVersionUpgrade == true AND target minor not vetted
+    ids: [TOP-04]
+    severity: LOW
+    message: "Auto minor upgrade may apply an unvetted minor at the next window"
+ASSESSMENT COMMANDS
+
+# Current engine/version, Serverless config, parameter group, Global DB membership
+aws rds describe-db-clusters --db-cluster-identifier {{CLUSTER}} --region {{REGION}} \
+  --query "DBClusters[0].{Engine:Engine,Version:EngineVersion,ClusterPG:DBClusterParameterGroup,Serverless:ServerlessV2ScalingConfiguration,GlobalId:GlobalClusterIdentifier,Members:DBClusterMembers}"
+
+# Valid upgrade targets + parameter group family for the CURRENT version
+aws rds describe-db-engine-versions --engine {{ENGINE}} --engine-version {{CURRENT_VERSION}} \
+  --region {{REGION}} \
+  --query "DBEngineVersions[0].{Family:DBParameterGroupFamily,ValidUpgradeTargets:ValidUpgradeTarget[].{Version:EngineVersion,IsMajor:IsMajorVersionUpgrade}}"
+
+# Target version parameter group family (for family-match check)
+aws rds describe-db-engine-versions --engine {{ENGINE}} --engine-version {{TARGET_VERSION}} \
+  --region {{REGION}} --query "DBEngineVersions[0].DBParameterGroupFamily"
+
+# Instance classes (target class availability)
+aws rds describe-orderable-db-instance-options --engine {{ENGINE}} --engine-version {{TARGET_VERSION}} \
+  --region {{REGION}} --query "OrderableDBInstanceOptions[].DBInstanceClass" --output text
+
+# Cluster parameters (check for removed/incompatible variables)
+aws rds describe-db-cluster-parameters --db-cluster-parameter-group-name {{CLUSTER_PG}} \
+  --region {{REGION}} --source user --query "Parameters[].{Name:ParameterName,Value:ParameterValue}"
+
+# Pending maintenance actions
+aws rds describe-pending-maintenance-actions --region {{REGION}} \
+  --query "PendingMaintenanceActions[?ResourceIdentifier=='{{CLUSTER_ARN}}']"
+
+# Recent snapshots (rollback point)
+aws rds describe-db-cluster-snapshots --db-cluster-identifier {{CLUSTER}} \
+  --snapshot-type manual --region {{REGION}} \
+  --query "DBClusterSnapshots[].{Id:DBClusterSnapshotIdentifier,Created:SnapshotCreateTime}"
+Manual in-database pre-checks (flag these to the operator)
+
+-- PostgreSQL: installed extensions vs target compatibility
+SELECT extname, extversion FROM pg_extension;
+-- PostgreSQL: active logical replication slots (must be empty to upgrade)
+SELECT slot_name, active FROM pg_replication_slots WHERE slot_type = 'logical';
+-- PostgreSQL: deprecated data types in use
+SELECT n.nspname, c.relname, a.attname, t.typname
+  FROM pg_attribute a JOIN pg_class c ON a.attrelid=c.oid
+  JOIN pg_type t ON a.atttypid=t.oid JOIN pg_namespace n ON c.relnamespace=n.oid
+  WHERE t.typname IN ('abstime','reltime','tinterval');
+
+-- Aurora MySQL: review the upgrade prechecks log after a dry-run/clone upgrade
+-- (upgrade-prechecks.log in the cluster's log exports)
+ASSESSMENT SCORING MATRIX
+Score Range	Rating	Meaning
+90-100	READY	Valid path, no blockers, snapshot + runbook in place
+70-89	MOSTLY READY	Minor warnings (stats loss, auto-minor) — proceed with runbook
+50-69	NEEDS PREP	Parameter/extension/config items to resolve first
+30-49	BLOCKED (fixable)	Hard blockers present but resolvable (slots, extensions, path)
+0-29	BLOCKED	Invalid path / Serverless v1->v2 direct / multiple hard blockers
+Scoring dimensions (25 pts each):
+Version path (25 pts): Valid direct target (+15); target class available (+5); not EOL/Extended Support (+5)
+
+Engine blockers (25 pts): No extension/precheck blockers (+15); no deprecated types/variables (+10)
+
+Config readiness (25 pts): Parameter group family matches target (+10); Serverless config correct (+8); Global DB coordination planned (+7)
+
+Safety & rollback (25 pts): Recent snapshot exists (+10); pending maintenance reconciled (+5); post-upgrade steps (ANALYZE, ALTER EXTENSION) planned (+10)
+
+UPGRADE RUNBOOK TEMPLATES
+PostgreSQL major version upgrade (sequenced)
+
+# 1. Snapshot (rollback point)
+aws rds create-db-cluster-snapshot --db-cluster-identifier {{CLUSTER}} \
+  --db-cluster-snapshot-identifier {{CLUSTER}}-pre-upgrade-{{DATE}}
+
+# 2. Resolve blockers (manual, per pre-checks):
+#    - Remove/consume logical replication slots
+#    - Drop/update incompatible extensions; fix deprecated data types
+#    - Create a target-family parameter group and set required params
+
+# 3. Test on a clone first (recommended)
+aws rds restore-db-cluster-to-point-in-time --source-db-cluster-identifier {{CLUSTER}} \
+  --db-cluster-identifier {{CLUSTER}}-upgrade-test --restore-type copy-on-write --use-latest-restorable-time
+
+# 4. Upgrade (maintenance window)
+aws rds modify-db-cluster --db-cluster-identifier {{CLUSTER}} \
+  --engine-version {{TARGET_VERSION}} \
+  --db-cluster-parameter-group-name {{TARGET_FAMILY_PG}} \
+  --allow-major-version-upgrade --apply-immediately
+
+# 5. Post-upgrade (IMMEDIATE):
+#    - Run ANALYZE (whole DB) to rebuild optimizer statistics
+#    - ALTER EXTENSION <ext> UPDATE; for pg_stat_statements and others
+Aurora MySQL Serverless v1 -> v2 migration (multi-step)
+
+# There is NO direct v1->v2 path. Sequence:
+# 1. Snapshot the v1 cluster
+aws rds create-db-cluster-snapshot --db-cluster-identifier {{V1_CLUSTER}} \
+  --db-cluster-snapshot-identifier {{V1_CLUSTER}}-pre-migrate
+
+# 2. Upgrade/restore to a provisioned Aurora MySQL version that supports v2
+#    (verify the migration-capable version via describe-db-engine-versions)
+
+# 3. Configure Serverless v2 scaling on the target cluster
+aws rds modify-db-cluster --db-cluster-identifier {{TARGET_CLUSTER}} \
+  --serverless-v2-scaling-configuration MinCapacity={{MIN_ACU}},MaxCapacity={{MAX_ACU}}
+
+# 4. Add a Serverless v2 instance to the cluster
+aws rds create-db-instance --db-instance-identifier {{TARGET}}-sv2-1 \
+  --db-instance-class db.serverless --engine aurora-mysql \
+  --db-cluster-identifier {{TARGET_CLUSTER}}
+
+# 5. Update application endpoints; validate; then decommission v1
+Global Database upgrade coordination
+
+# Upgrade requires per-region coordination and version parity for DR.
+# Custom parameter groups must be re-applied per region post-upgrade.
+# Verify version parity across members before/after:
+aws rds describe-global-clusters --global-cluster-identifier {{GLOBAL_ID}} \
+  --query "GlobalClusters[0].GlobalClusterMembers[].DBClusterArn"
+REPORT OUTPUT FORMAT
+
+# Aurora Upgrade Readiness Report
+**Cluster:** {{CLUSTER}} | **Engine:** {{ENGINE}} {{CURRENT_VERSION}} -> {{TARGET_VERSION}}
+**Deployment:** {{Provisioned | Serverless v1 | Serverless v2}} | **Region:** {{REGION}} | **Date:** {{DATE}}
+
+## Readiness Verdict: {{SCORE}}/100 ({{RATING}})
+
+## Version Path
+|
+ Check 
+|
+ Result 
+|
+|
+-------
+|
+--------
+|
+|
+ Valid direct upgrade target 
+|
+|
+|
+ Multi-hop required 
+|
+|
+|
+ Target instance class available 
+|
+|
+|
+ Extended Support status 
+|
+|
+
+## Blockers Detected
+|
+ Severity 
+|
+ ID 
+|
+ Blocker 
+|
+ Resolution 
+|
+ Blocking? 
+|
+|
+----------
+|
+-----
+|
+---------
+|
+-----------
+|
+-----------
+|
+
+## Manual Pre-Checks Required (in-database)
+- [ ] Extensions compatibility (`SELECT * FROM pg_extension;`)
+- [ ] Logical replication slots empty
+- [ ] Deprecated data types absent
+- [ ] MySQL upgrade-prechecks.log reviewed (clone dry-run)
+
+## Sequenced Runbook
+### Pre-upgrade (resolve blockers + snapshot)
+### Upgrade (maintenance window)
+### Post-upgrade (ANALYZE, ALTER EXTENSION, validation)
+
+## Rollback Plan
+- Snapshot: {{snapshot-id}} | Restore command ready: {{yes/no}}