cobalt-core
$
npx mdskill add notque/vexjoy-agent/cobalt-coreManages KVM hypervisor metrics and OpenStack compute infrastructure
- Solves hypervisor monitoring and compute node configuration tasks
- Uses KVM exporters, libvirt, and Prometheus for metrics collection
- Analyzes NUMA, cgroups, and hypervisor signals from reference patterns
- Provides actionable scripts and agent commands for infrastructure tuning
SKILL.md
.github/skills/cobalt-coreView on GitHub ↗
---
name: cobalt-core
description: "Cobalt Core infrastructure knowledge: KVM exporters, hypervisor tooling, OpenStack compute."
agent: kubernetes-helm-engineer
user-invocable: true
argument-hint: "<cobaltcore topic or repo name>"
allowed-tools:
- Read
- Grep
- Glob
- Bash
- Agent
- Edit
- Write
routing:
triggers:
- "cobalt core"
- "cobaltcore"
- "kvm-exporter"
- "kvm exporter"
- "hypervisor metrics"
- "libvirt exporter"
- "cloud hypervisor"
category: infrastructure
pairs_with:
- go-patterns
- prometheus-grafana-engineer
- kubernetes-helm-engineer
---
# Cobalt Core
Domain skill for the [cobaltcore-dev](https://github.com/cobaltcore-dev) project family — SAP Converged Cloud infrastructure components for KVM hypervisor management, metrics collection, and compute-node tooling.
## Reference Loading Table
| Signal | Reference | Size |
|--------|-----------|------|
| kvm-exporter, metrics, prometheus, libvirt, hypervisor, collector, scrape, steal time, NUMA, cgroups, cloud hypervisor | `references/kvm-exporter.md` | ~800 lines |
| goroutine, concurrency, semaphore, TryLock, sync.Map, race condition, socket exhaustion, scrape overlap, ClearScrapeCache | `references/concurrency-patterns.md` | ~200 lines |
| test, mock, moq, unit test, E2E, Kind cluster, race detector, interface_mock_gen, test-metrics.sh | `references/testing-patterns.md` | ~200 lines |
**Load greedily.** If the user's question touches any signal keyword, load the matching reference before responding. Multiple signals matching = load all matching references.
---
## Phase 1: IDENTIFY
Determine which cobaltcore component the user is asking about.
| Component | Repository | Reference |
|-----------|-----------|-----------|
| KVM Exporter | `cobaltcore-dev/kvm-exporter` | `references/kvm-exporter.md` |
If the component is not listed, tell the user no reference exists yet and offer to analyze the repo.
**Gate**: Component identified. Reference loaded. Proceed to Phase 2.
---
## Phase 2: RESPOND
Use loaded reference knowledge to answer the user's question. The references contain:
- Architecture and data flow diagrams
- Complete metric catalogs with types, labels, and descriptions
- Configuration options and environment variables
- Deployment models (Helm, DaemonSet, container specs)
- Code patterns (concurrency, caching, error handling)
- Testing strategies (unit mocks, E2E with Kind clusters)
- Alerting rules and operational concerns
For implementation questions involving Go code, pair with the `go-patterns` skill for language-specific patterns. For Prometheus/Grafana questions, pair with `prometheus-grafana-engineer`. For Kubernetes deployment questions, pair with `kubernetes-helm-engineer`.
**Gate**: Question answered with reference-backed specifics, not generic advice.
---
## Phase 3: EXTEND
When the user wants to add a new cobaltcore repo:
1. Analyze the repo systematically (README, go.mod, key source files, Dockerfile, Helm chart)
2. Create a new reference file at `references/{repo-name}.md`
3. Update the Reference Loading Table in this SKILL.md
4. Update the component table in Phase 1
Follow the structure established in `references/kvm-exporter.md` for consistency.
More from notque/vexjoy-agent
- adr-consultationMulti-agent consultation for architecture decisions.
- agent-comparisonA/B test agent variants for quality and token cost.
- agent-evaluationEvaluate agents and skills for quality and standards compliance.
- architecture-deepeningProactive architecture improvement: find shallow modules, propose deepening opportunities, design conversation.
- auto-dreamBackground memory consolidation and learning graduation — overnight knowledge lifecycle.
- bluesky-readerRead public Bluesky feeds via AT Protocol API.
- code-cleanupDetect stale TODOs, unused imports, and dead code.
- code-lintingRun Python (ruff) and JavaScript (Biome) linting.
- codebase-analyzerStatistical rule discovery from Go codebase patterns.
- codebase-overviewSystematic codebase exploration and architecture mapping.