kubernetes-sigs

Create (or reuse) a single kind cluster for Kueue e2e development. Use when a user asks to spin up an e2e cluster, get a local kind environment ready for e2e tests, or debug/iterate against a single-cluster e2e suite (baseline, extended, sequential, TAS,…

Review MultiKueue changes for unnecessary reads from remote worker clusters.

Review Kueue enhancement proposals (KEPs) for project-wide design concerns and readability. Use when reviewing a new or updated KEP.

Write or revise a Kueue enhancement proposal (KEP) so maintainers can review it quickly. Use when drafting a new KEP, updating an existing one, or before opening or pushing a KEP pull request.

Open or update Kueue pull requests and issues following the required templates, labels, titles, and AI disclosure rules.

Run Kueue's complete verification suite before pushing changes or opening or updating a pull request. Use as final validation to catch CI failures, including stale generated files.

Flag unit or integration test function names that do not reflect the function or behavior under test.

Review a diff for naming precision, convention drift, reinvented helpers, log-verbosity mistakes, misaligned test names, and typos.

Raise Go modules to caller-supplied minimum fixed versions from CVE/GO findings in any format, per tracked module root, sync go.mod/go.sum and root vendor/, audit the source module graphs, run the given checks, and publish a PR, all enforced by one script.…

Build cloud-provider-azure container images through the repo Makefile with explicit IMAGE_TAG and IMAGE_REGISTRY inputs, optional make flag overrides, and opt-in bounded Docker or Podman retries. Use when the user wants to build CCM, CNM, health-probe-proxy,…

Diagnose and unblock failed Dependabot pull requests in cloud-provider-azure by closing Kubernetes minor-version dependency bumps, classifying CI failures, syncing Go modules, retesting quota-flaked e2e jobs, and updating PR status. Use when a Dependabot PR…

Scan a built container image with Trivy, classify fixable Go-module and base-image CVEs, apply dependency and Dockerfile fixes, and verify the result with file checks and an optional image rescan. Use when the user wants to fix CVEs in a container image, scan…

Orchestrate end-to-end CVE remediation for the Linux CCM, CNM, and health-probe-proxy images on cloud-provider-azure master or a release-X.Y branch, including builds, repeated Trivy verification, checkpoint commits, cleanup, validation, push, and pull-request…

Synchronize all tracked Go modules with the go-mod-consistency CI and refresh the root vendor tree. Use when go-mod-consistency fails, dependency bumps leave go.mod or go.sum stale, or the user asks to unblock module consistency, tidy all modules, or update…

Fetch and analyze Prow e2e pipeline failures for cloud-provider-azure. Downloads build logs, JUnit reports, and node-level artifacts from GCS, extracts failed tests, parses Ginkgo results, and surfaces cluster node info. Use when debugging a failing Prow e2e…

Generate or update the documentation-site release note for a given tag, commit it on a branch, push it to a writable remote, and open a GitHub PR to the docs branch. Use when the user wants to create or update release notes, generate a changelog for a tag, or…

Guide compatible API, CRD, opaque device configuration, ResourceSlice contract, Helm value, kubelet-plugin checkpoint schema, and feature-gate changes in the DRA Driver for NVIDIA GPUs. Use when a user mentions CRDs, API fields, ResourceSlice attributes or…

Guide local builds, formatting, linting, code generation, unit and integration tests, BATS, end-to-end coverage, dependency checks, PR preparation, and CI debugging for the DRA Driver for NVIDIA GPUs. Use when a user asks how to test or validate a change,…

Explain, design, and troubleshoot NVIDIA ComputeDomain and IMEX workflows across MNNVL topology, daemon lifecycle, clique readiness, DRA channel allocation, and workload injection. Use when asked about ComputeDomains, IMEX, MNNVL, GB200 or GB300 systems,…

Explain, design, and troubleshoot NVIDIA GPU allocation through Kubernetes DRA, including ResourceClaims, ResourceSlices, scheduler selection, kubelet Prepare, CDI injection, MIG, and VFIO. Use when asked about GPU claims, MIG allocation, VFIO passthrough,…

Guide direct and GPU Operator-managed deployment, kind demos, sample manifest authoring, and quickstart validation for the NVIDIA DRA driver. Use when a user asks to install or demo the driver, create a resource.k8s.io ResourceClaim or ResourceClaimTemplate…

Guide triage of install, allocation, scheduling, runtime injection, plugin, controller, daemon, metric, and upgrade failures in the NVIDIA DRA Driver for GPUs. Use when a user provides logs, symptoms, failed pods, broken claims, missing devices, cluster…

Orient agents to the DRA Driver for NVIDIA GPUs repository by explaining what it does, mapping its runtime components and request flows, locating implementation, documentation, tests, and deployment code, and identifying source ownership and generated…

Draft or review maintainer-ready design proposals, architecture decision records, implementation plans, feature-gate plans, and scope analyses for the DRA Driver for NVIDIA GPUs. Use when a user asks for a proposal, design review, ADR, feature gate plan,…

Explain WHY a model scored low on an eval — read the judge's reasons and the agent's trajectory, compare against the task rubric, and produce a rationale (capability gap vs. task/rubric problem). Distinct from infrastructure failures. Invoke when someone asks…

Use when the user adds or changes a benchmark TASK and asks to review it — e.g. "review this new task", "is this task parallel-safe", "review this task.yaml + stack", "will this task collide under the matrix", "check my new tf/prebuilt stack". Reviews a newly…

Validate that a NEW eval is correct and runnable by running it and iterating in a loop until green — fixing infra/config/task bugs (not the model's score), with opt-in hands-off and unlimited/self-healing modes; invoke when authoring or vetting a task before…

Keep the docs current after a code change — map changed code areas to the docs that describe them, update those docs in place, and resolve known-issues rows the code has verifiably fixed. Invoke after editing the pipeline, adding a model/agent/task/metric,…

Run ONE evaluation — a single Task × Model × AgentConfig — end to end, local or on the bastion; invoke when the user wants to run, evaluate, or "kick off" a single eval combo (e.g. "run opa-remediation on gemini-3.1-pro-preview", "evaluate this task once and…

Run a Task × Model × AgentConfig MATRIX of evals in parallel — local or on the bastion, with opt-in hands-off and unlimited/self-healing modes; invoke when the user wants a matrix, a comparison across models/configs, or to "run these evals in parallel" /…

Use when the user asks for a CODE review of devops-bench changes — e.g. "review this PR", "review my changes", "review the working tree", "is this harness/deployer/metric change sound". Reviews a PR (number/URL) or the current working tree and returns ranked…

Discover and remove cloud or local resources leaked by aborted or failed eval runs — stale per-run state, leftover kind clusters and their node containers, stuck harness processes, and orphaned GKE clusters, node service accounts, secrets, VPCs and their…

Guides the agent to follow Kubernetes API conventions for OSS standards.

An MCP server skill for managing Kubernetes sandboxes. Enables creating, executing commands, managing files, and terminating instances via the official kubernetes-sigs/agent-sandbox MCP server.

Enforces project-specific development rules and conventions.

Analyze the repo's unit and E2E tests and propose rebalancing toward a test pyramid — which E2E tests (or assertions inside them) can be covered by unit tests, which unit-level gaps genuinely need E2E coverage, and where coverage is duplicated. Use when the…

Diagnose and fix flaky tests tracked as open kind/flake issues in kubernetes-sigs/agent-sandbox — reproduce the flake, apply a minimal fix, and open a PR linking the issue. Use when asked to fix a flake, work through kind/flake issues, or when run on a…

Triage open GitHub issues for kubernetes-sigs/agent-sandbox by mapping them to roadmap.md and assigning k8s priority labels + Kanban Priority (P0–P4) on Project

Bumps the Go version to the latest release across go.mod, tools.mod, and Dockerfiles.

Standards for CLI command and flag descriptions in Kubebuilder. Use when writing or reviewing CLI flags, commands, or help text.

Guidelines for maintaining the Kubebuilder Helm plugin (helm/v2-alpha) in pkg/plugins/optional/helm/v2alpha/. Use when working on Helm plugin code, templates, or chart generation.

Standards for Kubebuilder documentation in docs/book/src/. Use when writing or reviewing documentation, tutorials, guides, or plugin docs.

Bump Go to a new minor or patch version across the entire project. Use when upgrading Go (e.g. from 1.24 to 1.25). Handles go.mod, Makefile, golangci-lint, GitHub Actions, netlify.toml, and cloudbuild.

Bump Kubernetes and Cluster API to new minor versions across the entire project. Use when upgrading k8s (e.g. from 1.34 to 1.35) and CAPI (e.g. from 1.12 to 1.13). Handles go.mod, Makefile, versions.mk, metadata, e2e configs, CRDs, and migration-guide-driven…

Build, deploy, and debug the alibaba-cloud-csi-driver on a test cluster. Use this skill when the user wants to build images, deploy via helm, check logs, restart components, or troubleshoot the CSI plugin/provisioner. Also trigger for "deploy csi", "check csi…

Initialize the local development environment for the alibaba-cloud-csi-driver project. Use this skill when the user wants to set up their dev environment, configure test cluster access, create image pull secrets, or write test-values.yaml. Also trigger when…

Stage and commit changes with a Kubernetes-style commit message

Create a GitHub pull request from the current branch

Creates a Kind cluster with Workload-Aware Scheduling (WAS) feature gates enabled, deploys JobSet with the JobSetWorkloadAwareSchedulingAPI feature gate turned on, and provides example YAMLs for verifying gang scheduling, topology-constrained workers, and…

Debug CAPZ (Cluster API Provider Azure) Kubernetes cluster failures. Covers live cluster inspection via kubectl, VM-level debugging via az CLI, Prow/GCS artifact analysis, and build log triage. Knows CAPZ template flavors, E2E test structure, addon deployment…