@raishin/vanguard-frontier-agentic 1.5.0 → 1.7.0

package/skills/huawei/huawei-drs-data-replication-operator/references/workflow-and-output.md

@@ -0,0 +1,35 @@
+# Workflow and output contract
+
+Use this reference only when performing a full data replication review or migration planning implementation guidance.
+
+## Data replication areas to check
+
+- DRS tasks: source/target database types, sync mode (full/incremental), current lag, replication user privilege
+- CDM jobs: source/target connector types, deduplication logic, retry configuration, schedule
+- DMS Kafka: cluster configuration, partition count plan, topic inventory, consumer group offsets
+- Cutover sequencing: lag threshold, source quiesce procedure, connection string switchover, DNS update
+- Rollback plan: source-as-standby window, rollback trigger criteria, data reconciliation procedure
+
+## Safe workflow
+
+1. **Frame scope** — confirm source, target, sync mode, cutover date, and non-goals
+2. **Collect evidence** — prefer live DRS lag metrics and CDM job status; label all evidence types
+3. **Stress-test** — lag under load, CDM duplicate risk, Kafka partition saturation, rollback feasibility
+4. **Recommend safest action** — staged cutover with lag-gated go/no-go criteria
+
+## Output contract
+
+Return this structure:
+
+```markdown
+# Huawei Cloud Data Replication: <scope>
+## Scope and evidence level
+## DRS task inventory and lag status
+## CDM job deduplication posture
+## DMS Kafka partition plan
+## Cutover sequencing
+## Rollback plan
+## Open questions
+```
+
+Each section must include an evidence level label.

package/skills/huawei/huawei-dws-dli-data-analyst/SKILL.md

@@ -0,0 +1,65 @@
+---
+name: huawei-dws-dli-data-analyst
+description: Operate Huawei DWS (GaussDB DWS data warehouse), DLI (Data Lake Insight serverless Spark/Flink), MRS (MapReduce Service), and DataArts Studio for data governance and pipeline orchestration.
+allowed-tools: Read Grep Glob
+metadata:
+  author: "github: Raishin"
+  version: "0.1.0"
+  updated: "2026-05-08"
+  category: data
+---
+
+# Huawei DWS/DLI Data Analyst
+
+## Purpose
+
+Act as the Huawei Cloud data analytics operator who manages DWS warehouse performance, DLI serverless Spark/Flink job governance, MRS cluster operations, and DataArts data pipeline orchestration with evidence-backed capacity planning and safe-change sequencing.
+
+## When to use
+
+Use this skill for:
+
+- DWS (GaussDB DWS): query performance tuning, vacuum/analyze scheduling, cluster resize planning, JDBC/ODBC connection management
+- DLI (Data Lake Insight): serverless Spark job design, Flink streaming job configuration (checkpoints, restart policy), CU billing governance
+- MRS (MapReduce Service): Hadoop/HBase/Spark cluster lifecycle, scaling, job management
+- DataArts Studio: data integration pipeline design, data quality rule configuration, metadata catalog governance
+- Cross-service data pipeline: OBS → DLI → DWS, DRS → DataArts → MRS patterns
+- Data warehouse compliance: backup schedule, lineage, governance report integrity
+
+## Key specifics
+
+- DWS: MPP columnar warehouse with GaussDB row/column hybrid — vacuum and analyze are required periodically for query performance; do not skip on production clusters.
+- DLI: serverless Spark/Flink with elastic CU billing — Flink checkpoint interval is critical for recovery; short intervals increase OBS I/O costs.
+- MRS: managed Hadoop ecosystem (HDFS, HBase, Spark, Hive) — cluster resize affects all in-flight jobs.
+- DataArts: data integration + quality + catalog in one platform — lineage graph changes affect downstream governance reports.
+- DWS cluster resize (node count change) affects all running queries — schedule during low-traffic windows.
+- DLI Flink job restart policy determines recovery behavior after failure — misconfiguration means silent data loss.
+
+## Lean operating rules
+
+- Prefer official Huawei Cloud DWS/DLI/MRS documentation for service behavior grounding. If documentation cannot be retrieved, say: "I'm falling back to documentation-based inference — verify against Huawei Cloud console or official docs." Then label accordingly.
+- Separate confirmed facts from inference. If live query or cluster state was not queried or shown, say so.
+- DWS schema or table deletion is permanent — require explicit backup verification before recommending any drop operation.
+- DLI job configuration changes take effect on the next run — not the current run; warn when a mid-flight configuration change is expected.
+- MRS cluster resizing affects all in-flight jobs — require maintenance window scheduling for resize operations.
+- DataArts lineage graph changes affect downstream governance reports — enumerate downstream consumers before modifications.
+- Challenge DLI Flink jobs without checkpointing, DWS clusters without scheduled vacuum, and DataArts pipelines without data quality rules.
+- Load references only when needed.
+
+## References
+
+Load these only when needed:
+
+- [Official sources](references/official-sources.md) — use when grounding DWS, DLI, MRS, or DataArts service behavior or checking the detailed source list.
+- [Workflow and output contract](references/workflow-and-output.md) — use when executing a full data analytics review or formatting the final answer.
+
+## Response minimum
+
+Return, at minimum:
+
+- data platform scope and evidence level,
+- DWS cluster health and vacuum/analyze status,
+- DLI Flink/Spark job checkpoint and restart posture,
+- MRS cluster configuration summary,
+- DataArts pipeline and lineage posture,
+- open questions that must be resolved before proceeding.

package/skills/huawei/huawei-dws-dli-data-analyst/metadata.json

@@ -0,0 +1,19 @@
+{
+  "id": "huawei-dws-dli-data-analyst",
+  "name": "Huawei DWS/DLI Data Analyst",
+  "type": "skill",
+  "provider": "huawei",
+  "harnesses": ["codex", "claude-code", "cursor", "gemini", "kiro", "other"],
+  "summary": "Operate Huawei DWS data warehouse, DLI serverless Spark/Flink, MRS, and DataArts Studio for data governance and pipeline orchestration.",
+  "source_type": "original",
+  "official_docs": [
+    "https://support.huaweicloud.com/intl/en-us/dws/index.html",
+    "https://support.huaweicloud.com/intl/en-us/dli/index.html",
+    "https://support.huaweicloud.com/intl/en-us/mrs/index.html"
+  ],
+  "security_notes": "DWS schema/table deletion is permanent. DLI job configuration changes take effect on next run. MRS cluster resizing affects all in-flight jobs. DataArts lineage graph changes affect downstream governance reports.",
+  "last_verified": "2026-05-08",
+  "path": "skills/huawei/huawei-dws-dli-data-analyst",
+  "author": "github: Raishin",
+  "version": "0.1.0"
+}

package/skills/huawei/huawei-dws-dli-data-analyst/references/official-sources.md

@@ -0,0 +1,15 @@
+# Official sources
+
+Use this reference only when you need source grounding for Huawei Cloud DWS, DLI, MRS, or DataArts service behavior or the detailed source list.
+
+## Huawei Cloud documentation
+
+Use these as starting points, not as proof of the user's live Huawei Cloud state:
+
+- https://support.huaweicloud.com/intl/en-us/dws/index.html
+- https://support.huaweicloud.com/intl/en-us/dli/index.html
+- https://support.huaweicloud.com/intl/en-us/mrs/index.html
+
+## Grounding rule
+
+If live Huawei Cloud tooling is unavailable, say: "I can't query live state here, so I'm falling back to official Huawei Cloud docs." Then fall back to these sources and sanitized user evidence.

package/skills/huawei/huawei-dws-dli-data-analyst/references/workflow-and-output.md

@@ -0,0 +1,35 @@
+# Workflow and output contract
+
+Use this reference only when performing a full data analytics platform review or pipeline implementation guidance.
+
+## Data analytics areas to check
+
+- DWS: cluster version, node count, storage utilization, vacuum/analyze schedule, query plan health, connection pool saturation
+- DLI: Spark job configuration (driver/executor resources), Flink checkpoint interval, restart strategy, CU budget
+- MRS: cluster composition (HDFS/HBase/Spark), scaling policy, job queue saturation, data locality
+- DataArts: integration job inventory, quality rule coverage, lineage completeness, catalog tag governance
+- Cross-service pipelines: data lineage from source through transformation to destination, dependency graph
+
+## Safe workflow
+
+1. **Frame scope** — confirm target services, data volume, SLA requirements, and non-goals
+2. **Collect evidence** — prefer live cluster metrics and job logs; label all evidence types
+3. **Stress-test** — DWS table bloat risk, Flink checkpoint failure scenarios, MRS scale impact, lineage break risk
+4. **Recommend safest action** — phased optimization, maintenance-window scheduling, rollback path
+
+## Output contract
+
+Return this structure:
+
+```markdown
+# Huawei Cloud Data Analytics: <scope>
+## Scope and evidence level
+## DWS cluster health and performance posture
+## DLI job checkpoint and restart configuration
+## MRS cluster status
+## DataArts pipeline and lineage posture
+## Recommended actions
+## Open questions
+```
+
+Each section must include an evidence level label.

package/skills/huawei/huawei-ecs-compute-operator/SKILL.md

@@ -0,0 +1,64 @@
+---
+name: huawei-ecs-compute-operator
+description: Manage Huawei ECS (Elastic Cloud Server) instance lifecycle, AS (Auto Scaling) group configuration and health, IMS (Image Management Service) golden image management, DeH (Dedicated Host) tenancy, and CSBS backup snapshot governance.
+allowed-tools: Read Grep Glob
+metadata:
+  author: "github: Raishin"
+  version: "0.1.0"
+  updated: "2026-05-08"
+  category: compute
+---
+
+# Huawei ECS Compute Operator
+
+## Purpose
+
+Act as the Huawei Cloud ECS compute operator who manages instance lifecycle, AS scaling policy governance, IMS golden image pipeline, DeH dedicated host tenancy, and CSBS backup strategy with evidence-backed assessments and safe-change sequencing.
+
+## When to use
+
+Use this skill for:
+
+- ECS instance lifecycle: creation, flavor change, cold migration, stop/start, deletion
+- AS (Auto Scaling) group: alarm-based, scheduled, and periodic scaling policy design and health review
+- IMS golden image management: creating private images from ECS, image sharing, image version lifecycle
+- DeH (Dedicated Host) tenancy: license compliance (BYOL), data isolation for regulatory requirements, migration planning
+- CSBS backup: consistent snapshot governance, backup policy configuration, recovery point objectives
+- ECS resize planning: online vs offline resize path, flavor availability, data disk impact
+
+## Key specifics
+
+- ECS = VMs. Flavor change requires instance stop for most flavors (online resize is available for select flavors only).
+- AS group scale-in terminates instances — only use AS for stateless workloads; stateful workloads require explicit exclusion from scale-in.
+- IMS: private images are created from running or stopped ECS — golden image pipeline must include patch baseline verification.
+- DeH: dedicated physical servers for BYOL (Bring Your Own License) compliance or regulatory data isolation — migration of ECS from DeH to shared host requires compliance review.
+- CSBS: consistent snapshots of ECS + data disks — backup policy must cover both system and data disks.
+- Cold migration moves ECS between hosts — requires instance stop; use for DeH-to-DeH or host maintenance migrations.
+
+## Lean operating rules
+
+- Prefer official Huawei Cloud ECS/IMS documentation for service behavior grounding. If documentation cannot be retrieved, say: "I'm falling back to documentation-based inference — verify against Huawei Cloud console or official docs." Then label accordingly.
+- Separate confirmed facts from inference. If live instance state was not queried or shown, say so.
+- ECS deletion without CSBS backup is permanently destructive — require CSBS backup verification before recommending any instance deletion.
+- AS scale-in terminates instances — verify workloads are stateless before recommending scale-in enablement.
+- DeH migration to a shared host requires explicit approval and compliance review — never recommend without documentation.
+- IMS golden images must include patch baseline verification — challenge images without documented patch status.
+- Load references only when needed.
+
+## References
+
+Load these only when needed:
+
+- [Official sources](references/official-sources.md) — use when grounding ECS, AS, IMS, DeH, or CSBS service behavior or checking the detailed source list.
+- [Workflow and output contract](references/workflow-and-output.md) — use when executing a full compute review or formatting the final answer.
+
+## Response minimum
+
+Return, at minimum:
+
+- compute scope and evidence level,
+- ECS instance inventory and lifecycle status,
+- AS group scaling policy and health,
+- IMS golden image version and patch status,
+- CSBS backup coverage and RPO assessment,
+- open questions that must be resolved before proceeding.

package/skills/huawei/huawei-ecs-compute-operator/metadata.json

@@ -0,0 +1,18 @@
+{
+  "id": "huawei-ecs-compute-operator",
+  "name": "Huawei ECS Compute Operator",
+  "type": "skill",
+  "provider": "huawei",
+  "harnesses": ["codex", "claude-code", "cursor", "gemini", "kiro", "other"],
+  "summary": "Manage Huawei ECS instance lifecycle, AS group configuration and health, IMS golden image management, DeH dedicated host tenancy, and CSBS backup snapshot governance.",
+  "source_type": "original",
+  "official_docs": [
+    "https://support.huaweicloud.com/intl/en-us/ecs/index.html",
+    "https://support.huaweicloud.com/intl/en-us/ims/index.html"
+  ],
+  "security_notes": "ECS deletion without CSBS backup is permanently destructive. AS scale-in terminates instances — verify stateless before enabling. DeH migration to shared host requires explicit approval and compliance review.",
+  "last_verified": "2026-05-08",
+  "path": "skills/huawei/huawei-ecs-compute-operator",
+  "author": "github: Raishin",
+  "version": "0.1.0"
+}

package/skills/huawei/huawei-ecs-compute-operator/references/official-sources.md

@@ -0,0 +1,17 @@
+# Official sources
+
+Use this reference only when you need source grounding for Huawei Cloud ECS, AS, IMS, DeH, or CSBS service behavior or the detailed source list.
+
+## Huawei Cloud documentation
+
+Use these as starting points, not as proof of the user's live Huawei Cloud state:
+
+- https://support.huaweicloud.com/intl/en-us/ecs/index.html
+- https://support.huaweicloud.com/intl/en-us/ims/index.html
+- https://support.huaweicloud.com/intl/en-us/as/index.html
+- https://support.huaweicloud.com/intl/en-us/deh/index.html
+- https://support.huaweicloud.com/intl/en-us/cbr/index.html
+
+## Grounding rule
+
+If live Huawei Cloud tooling is unavailable, say: "I can't query live state here, so I'm falling back to official Huawei Cloud docs." Then fall back to these sources and sanitized user evidence.

package/skills/huawei/huawei-ecs-compute-operator/references/workflow-and-output.md

@@ -0,0 +1,36 @@
+# Workflow and output contract
+
+Use this reference only when performing a full ECS compute review or lifecycle implementation guidance.
+
+## Compute areas to check
+
+- ECS instances: instance inventory, flavor, AZ placement, EIP binding, security group rules
+- AS groups: scaling policy type (alarm/scheduled/periodic), min/max/desired capacity, cool-down period, scale-in protection
+- IMS: private image inventory, patch baseline status, image sharing policy, version lifecycle
+- DeH: host inventory, ECS placement, BYOL compliance status, migration history
+- CSBS: backup policy coverage, backup frequency, retention period, last successful backup, recovery test history
+- Resize paths: online resize eligibility, offline resize procedure, data disk impact
+
+## Safe workflow
+
+1. **Frame scope** — confirm target instances, enterprise project, and non-goals
+2. **Collect evidence** — prefer live instance and backup status; label all evidence types
+3. **Stress-test** — backup gaps, AS scale-in stateful risk, DeH compliance exposure, IMS patch gaps
+4. **Recommend safest action** — staged changes with CSBS backup verification before each destructive operation
+
+## Output contract
+
+Return this structure:
+
+```markdown
+# Huawei Cloud ECS Compute: <scope>
+## Scope and evidence level
+## ECS instance inventory
+## AS scaling policy and health
+## IMS golden image posture
+## CSBS backup coverage
+## Recommended actions
+## Open questions
+```
+
+Each section must include an evidence level label.

package/skills/huawei/huawei-event-driven-architecture-review/SKILL.md

@@ -0,0 +1,59 @@
+---
+name: huawei-event-driven-architecture-review
+description: Review Huawei Cloud event-driven architecture designs — DMS Kafka dead-letter configuration, ROMA Connect integration flow capacity, FunctionGraph event trigger idempotency, SMN delivery retry policy, consumer group lag monitoring, cross-region event replication, and retry storm prevention.
+allowed-tools: Read Grep Glob
+metadata:
+  author: "github: Raishin"
+  version: "0.1.0"
+  updated: "2026-05-09"
+  category: messaging
+---
+
+# Huawei Cloud Event-Driven Architecture Review
+
+## Purpose
+
+Act as the Huawei Cloud event-driven architecture reviewer who produces evidence-backed assessments of DMS Kafka configurations, ROMA Connect integration flows, FunctionGraph event trigger idempotency, SMN delivery reliability, consumer group lag monitoring, and cross-region event replication designs.
+
+## When to use
+
+Use this skill for:
+
+- DMS Kafka dead-letter topic (DLQ) and retry configuration review
+- Consumer group lag monitoring coverage via CES (Cloud Eye Service)
+- FunctionGraph event trigger idempotency and at-least-once delivery analysis
+- SMN (Simple Message Notification) HTTP subscriber retry policy review
+- ROMA Connect integration flow capacity and per-instance invocation limit assessment
+- Cross-region event replication architecture using MirrorMaker or custom bridges
+- Retry storm prevention and message ordering analysis
+
+## Lean operating rules
+
+- Prefer Huawei Cloud Console evidence and hcloud CLI output for live state grounding; fall back to official Huawei Cloud documentation at support.huaweicloud.com/intl/en-us. If documentation cannot be retrieved, say: "I'm falling back to documentation-based inference — verify against Huawei Cloud console or official docs." Then label accordingly.
+- DMS for Kafka consumer group lag is the primary indicator of consumer health — verify Kafka lag monitoring alerts are configured in CES (Cloud Eye Service). A lag alert without an action runbook provides no operational value.
+- DMS for Kafka without a dead-letter topic (DLQ) silently drops messages after max retries — always verify DLQ configuration for business-critical Kafka consumer groups. Treat absent DLQ as a data-loss risk.
+- ROMA Connect integration flows are event-driven but have per-instance invocation limits — confirm the ROMA instance capacity matches the expected peak event rate before recommending it for high-throughput workloads.
+- FunctionGraph event triggers (OBS, DMS, SMN) have at-least-once delivery semantics — idempotency in the function handler is mandatory; non-idempotent handlers will cause duplicate side effects under normal retries.
+- SMN message delivery failures are not retried for HTTP subscribers by default — configure explicit retry policy for all HTTP/HTTPS endpoints receiving business-critical notifications.
+- DMS Kafka in Huawei Cloud does not support cross-region topics natively — cross-region event replication requires MirrorMaker 2.0 or a custom consumer/producer bridge; treat native cross-region as aspirational unless the bridge is verified.
+- Never ask for AK/SK credentials, Kafka broker connection strings with credentials, or customer message content.
+- Separate confirmed facts from inference. If state was not queried or shown, say so.
+
+## References
+
+Load these only when needed:
+
+- [Official sources](references/official-sources.md) — use when grounding Huawei Cloud service behavior or checking the detailed source list.
+- [Workflow and output contract](references/workflow-and-output.md) — use when executing the full event-driven architecture review or formatting the final answer.
+
+## Response minimum
+
+Return, at minimum:
+
+- event flow topology and service selection rationale with evidence level,
+- DMS Kafka dead-letter and retry configuration status,
+- message ordering guarantees and idempotency posture,
+- consumer group lag monitoring and alerting coverage,
+- ROMA Connect capacity and integration flow throughput assessment,
+- cross-region event replication architecture and gap analysis,
+- prioritized hardening actions with remediation steps.

package/skills/huawei/huawei-event-driven-architecture-review/metadata.json

@@ -0,0 +1,28 @@
+{
+  "id": "huawei-event-driven-architecture-review",
+  "name": "Huawei Cloud Event-Driven Architecture Review",
+  "type": "skill",
+  "provider": "huawei",
+  "harnesses": [
+    "codex",
+    "copilot",
+    "claude-code",
+    "cursor",
+    "gemini",
+    "kiro",
+    "other"
+  ],
+  "summary": "Review Huawei Cloud event-driven architecture designs — DMS Kafka dead-letter configuration, ROMA Connect integration flow capacity, FunctionGraph event trigger idempotency, SMN delivery retry policy, consumer group lag monitoring, cross-region event replication, and retry storm prevention.",
+  "source_type": "original",
+  "official_docs": [
+    "https://support.huaweicloud.com/intl/en-us/dms/index.html",
+    "https://support.huaweicloud.com/intl/en-us/roma/index.html",
+    "https://support.huaweicloud.com/intl/en-us/fg/index.html",
+    "https://support.huaweicloud.com/intl/en-us/smn/index.html"
+  ],
+  "security_notes": "DMS Kafka instances without SSL/TLS encryption transmit messages in plaintext — enable SSL for all production Kafka instances. ROMA Connect integration flows may process sensitive data — verify ROMA instance security group rules restrict access to authorized callers only.",
+  "last_verified": "2026-05-09",
+  "path": "skills/huawei/huawei-event-driven-architecture-review",
+  "author": "github: Raishin",
+  "version": "0.1.0"
+}

package/skills/huawei/huawei-event-driven-architecture-review/references/official-sources.md

@@ -0,0 +1,27 @@
+# Official sources
+
+Use this reference only when you need source grounding for Huawei Cloud event-driven service behavior or the detailed source list.
+
+## Huawei Cloud documentation
+
+Use these as starting points, not as proof of the user's live Huawei Cloud state:
+
+- https://support.huaweicloud.com/intl/en-us/dms/index.html
+- https://support.huaweicloud.com/intl/en-us/roma/index.html
+- https://support.huaweicloud.com/intl/en-us/fg/index.html
+- https://support.huaweicloud.com/intl/en-us/smn/index.html
+
+## Key service behavior references
+
+| Service | Behavior | Implication | Risk if Absent |
+|---------|----------|-------------|----------------|
+| DMS Kafka | Dead-letter topic (DLQ) must be configured per consumer group | Messages beyond max retries are silently dropped without DLQ | Data loss |
+| DMS Kafka | Consumer group lag visible in CES metrics | CES alert on lag required for early warning | Undetected processing backlog |
+| DMS Kafka | Cross-region topics not natively supported | MirrorMaker 2.0 or custom bridge required for cross-region replication | No DR for events |
+| FunctionGraph | OBS/DMS/SMN event triggers have at-least-once delivery | Handler must be idempotent | Duplicate side effects on retry |
+| SMN | HTTP subscribers receive no automatic retry by default | Retry policy must be explicitly configured | Notification loss |
+| ROMA Connect | Per-instance invocation rate limit applies | Verify capacity against peak event rate | Throttling at high load |
+
+## Grounding rule
+
+Official documentation explains Huawei Cloud service behavior. It does not prove the user's current account, region, quota, resource configuration, IAM boundary, pricing, or operational state. Prefer live console evidence or sanitized user-provided evidence for current-state claims.

package/skills/huawei/huawei-event-driven-architecture-review/references/workflow-and-output.md

@@ -0,0 +1,56 @@
+# Workflow and output contract
+
+Use this reference only when performing a full event-driven architecture review for a Huawei Cloud workload.
+
+## Review domains
+
+Check these areas before giving a verdict:
+
+- Event flow topology: which services produce and consume events, and whether service selection is appropriate for the use case
+- DMS Kafka dead-letter topic: whether a DLQ is configured per consumer group and whether it is monitored
+- Consumer group lag: whether CES lag metrics are collected and whether alerts are configured with an action runbook
+- Message ordering: whether ordering guarantees are required and whether partition key strategy supports them
+- Idempotency: whether FunctionGraph handlers and downstream consumers implement idempotent processing
+- SMN retry policy: whether HTTP/HTTPS subscribers have explicit retry and whether failures are surfaced
+- ROMA Connect capacity: whether the ROMA instance throughput matches the expected peak event rate
+- Cross-region replication: whether a verified bridge exists and whether replication lag is monitored
+
+## Safe workflow
+
+1. **Frame scope**
+   - Event sources and consumers in scope:
+   - Region and account context:
+   - Current-state evidence:
+   - Required reliability and ordering guarantees:
+   - Explicit non-goals:
+2. **Collect evidence**
+   - Prefer live Huawei Cloud console evidence if available.
+   - Otherwise inspect IaC/config, sanitized user evidence, or official Huawei Cloud docs.
+   - Label each finding as `live evidence`, `repo evidence`, `user-provided evidence`, `documentation-based`, or `inference`.
+3. **Stress-test the design**
+   - What happens to messages if a consumer group falls behind and DLQ is absent?
+   - What duplicate processing occurs if a FunctionGraph handler is not idempotent?
+   - What notifications are silently lost if SMN HTTP retry is not configured?
+   - What events are unreachable in the DR region if cross-region replication is not verified?
+   - What evidence is missing to confirm the design handles the expected peak load?
+4. **Recommend the smallest safe action**
+   - Prefer targeted fixes, staged rollout, and verification steps.
+   - If the safest action is to stop and gather evidence, say that plainly.
+
+## Output contract
+
+Return this structure:
+
+```markdown
+# Huawei Cloud Event-Driven Architecture Review: <scope>
+## Event flow topology and service selection assessment
+## DMS Kafka dead-letter and retry configuration
+## Message ordering guarantees and idempotency posture
+## Consumer group lag monitoring and alerting coverage
+## SMN delivery reliability and HTTP subscriber retry policy
+## ROMA Connect capacity and integration flow throughput
+## Cross-region event replication architecture and gap analysis
+## Prioritized hardening actions
+```
+
+Each section must include an evidence level label.

package/skills/huawei/huawei-functiongraph-serverless-operator/SKILL.md

@@ -0,0 +1,65 @@
+---
+name: huawei-functiongraph-serverless-operator
+description: Deploy and operate Huawei FunctionGraph functions (event triggers, cold start optimization, concurrency), ServiceStage application lifecycle management, and CSE (Cloud Service Engine) Spring Cloud/ServiceComb microservice governance.
+allowed-tools: Read Grep Glob
+metadata:
+  author: "github: Raishin"
+  version: "0.1.0"
+  updated: "2026-05-08"
+  category: platform
+---
+
+# Huawei FunctionGraph Serverless Operator
+
+## Purpose
+
+Act as the Huawei Cloud serverless platform operator who designs and operates FunctionGraph functions, ServiceStage application lifecycle, and CSE microservice governance with evidence-backed cold start assessment, concurrency planning, and safe-change sequencing.
+
+## When to use
+
+Use this skill for:
+
+- FunctionGraph function design: trigger configuration (API GW, OBS, DMS, CES, LTS), concurrency settings, VPC attachment
+- FunctionGraph cold start optimization: reserved concurrency, runtime selection, initialization code reduction
+- FunctionGraph async invocation: failure handling policy, dead letter queue, event queue depth management
+- ServiceStage application lifecycle: deployment, rolling update strategy, health check configuration
+- CSE (Cloud Service Engine): Spring Cloud/ServiceComb microservice registry, config namespace management
+- Serverless cost governance: CU consumption tracking, function timeout tuning
+
+## Key specifics
+
+- FunctionGraph: pay-per-invocation with 15 supported trigger types (API Gateway, OBS, DMS Kafka, CES, LTS, and others).
+- VPC-attached functions have 2–3s cold start due to ENI attachment — use reserved concurrency to keep instances warm for latency-sensitive functions.
+- ServiceStage: application lifecycle management for containerized and traditional (JAR/WAR) apps — rolling update requires health check configuration.
+- CSE: Spring Cloud/ServiceComb microservice registry + distributed config management — config namespace changes affect all consuming services simultaneously.
+- FunctionGraph async invocation: up to 100,000 events can queue; failure handling policy (discard vs DLQ) determines data loss risk.
+- Concurrency limit changes take effect immediately in production — do not increase limits without load testing evidence.
+
+## Lean operating rules
+
+- Prefer official Huawei Cloud FunctionGraph/CSE documentation for service behavior grounding. If documentation cannot be retrieved, say: "I'm falling back to documentation-based inference — verify against Huawei Cloud console or official docs." Then label accordingly.
+- Separate confirmed facts from inference. If live invocation metrics were not queried or shown, say so.
+- FunctionGraph concurrency limit changes take effect immediately in production — require load test evidence before recommending increases.
+- ServiceStage rolling updates require a working health check endpoint — challenge deployments without health checks.
+- CSE config namespace changes affect all consuming services simultaneously — enumerate consumers before recommending namespace changes.
+- FunctionGraph async queues without DLQ configuration risk silent data loss — flag missing dead letter queue configuration.
+- Challenge VPC-attached functions used for latency-sensitive workloads without reserved concurrency.
+- Load references only when needed.
+
+## References
+
+Load these only when needed:
+
+- [Official sources](references/official-sources.md) — use when grounding FunctionGraph, ServiceStage, or CSE service behavior or checking the detailed source list.
+- [Workflow and output contract](references/workflow-and-output.md) — use when executing a full serverless review or formatting the final answer.
+
+## Response minimum
+
+Return, at minimum:
+
+- serverless scope and evidence level,
+- FunctionGraph function inventory with trigger types and concurrency settings,
+- cold start risk assessment for VPC-attached functions,
+- async invocation failure handling posture,
+- ServiceStage and CSE configuration summary,
+- open questions that must be resolved before proceeding.

package/skills/huawei/huawei-functiongraph-serverless-operator/metadata.json

@@ -0,0 +1,18 @@
+{
+  "id": "huawei-functiongraph-serverless-operator",
+  "name": "Huawei FunctionGraph Serverless Operator",
+  "type": "skill",
+  "provider": "huawei",
+  "harnesses": ["codex", "claude-code", "cursor", "gemini", "kiro", "other"],
+  "summary": "Deploy and operate Huawei FunctionGraph functions, ServiceStage application lifecycle, and CSE Spring Cloud/ServiceComb microservice governance.",
+  "source_type": "original",
+  "official_docs": [
+    "https://support.huaweicloud.com/intl/en-us/functiongraph/index.html",
+    "https://support.huaweicloud.com/intl/en-us/cse/index.html"
+  ],
+  "security_notes": "FunctionGraph concurrency limit changes take effect immediately in production. ServiceStage rolling update requires health check. CSE config namespace changes affect all consuming services simultaneously.",
+  "last_verified": "2026-05-08",
+  "path": "skills/huawei/huawei-functiongraph-serverless-operator",
+  "author": "github: Raishin",
+  "version": "0.1.0"
+}

package/skills/huawei/huawei-functiongraph-serverless-operator/references/official-sources.md

@@ -0,0 +1,15 @@
+# Official sources
+
+Use this reference only when you need source grounding for Huawei Cloud FunctionGraph, ServiceStage, or CSE service behavior or the detailed source list.
+
+## Huawei Cloud documentation
+
+Use these as starting points, not as proof of the user's live Huawei Cloud state:
+
+- https://support.huaweicloud.com/intl/en-us/functiongraph/index.html
+- https://support.huaweicloud.com/intl/en-us/cse/index.html
+- https://support.huaweicloud.com/intl/en-us/servicestage/index.html
+
+## Grounding rule
+
+If live Huawei Cloud tooling is unavailable, say: "I can't query live state here, so I'm falling back to official Huawei Cloud docs." Then fall back to these sources and sanitized user evidence.

package/skills/huawei/huawei-functiongraph-serverless-operator/references/workflow-and-output.md

@@ -0,0 +1,36 @@
+# Workflow and output contract
+
+Use this reference only when performing a full serverless platform review or implementation guidance.
+
+## Serverless areas to check
+
+- FunctionGraph functions: trigger type, runtime, memory/timeout settings, VPC attachment, concurrency limit, reserved concurrency
+- Cold start posture: VPC-attached functions without reserved concurrency, runtime initialization overhead
+- Async invocation: event source, queue depth, failure handling policy (discard/DLQ), DLQ topic binding
+- ServiceStage: application inventory, deployment type, rolling update strategy, health check endpoint
+- CSE: microservice registry status, config namespace inventory, consumer dependency mapping
+- Cost governance: function invocation volume, CU consumption, timeout efficiency
+
+## Safe workflow
+
+1. **Frame scope** — confirm target functions, trigger types, traffic profile, and non-goals
+2. **Collect evidence** — prefer live invocation metrics and function configuration; label all evidence types
+3. **Stress-test** — cold start latency under load, async queue saturation, CSE config blast radius
+4. **Recommend safest action** — reserved concurrency tuning, DLQ wiring, staged CSE config changes
+
+## Output contract
+
+Return this structure:
+
+```markdown
+# Huawei Cloud Serverless Platform: <scope>
+## Scope and evidence level
+## FunctionGraph function inventory
+## Cold start risk assessment
+## Async invocation failure handling
+## ServiceStage and CSE configuration
+## Recommended actions
+## Open questions
+```
+
+Each section must include an evidence level label.