Judgment layer for an **agent that acts on Confluence** (creates, organises, governs content) and, second, for a **human authoring pages** — the same structure serves both: what makes a page scannable makes it retrievable.

Run and upgrade a **self-managed GitLab on the official Helm chart** without unplanned downtime. The hard part is never `helm upgrade`. It is that the chart changes defaults the site never set, ships prerequisites written for a configuration the site does not run, and removed its own bundled databas

Host-side tuning + audit for Linux GPU inference servers. Sits *beneath* any inference framework (vLLM, sglang, TensorRT-LLM, llama.cpp).

Plan, sequence, and de-risk a **community-edition Harvester HCI** upgrade across a real fleet — typically off an **EOL line** (1.5.x) up to the latest *settled* stable. The hard part is almost never the ISO/UI upgrade click. It is the **coordination**: Harvester is one bundle whose embedded RKE2,

KEDA extends Kubernetes HPA with event-driven scalers (queues, cron, Prometheus, etc.) and owns the `0 ↔ 1` transition so workloads can truly scale to zero. The skill covers three CRDs (`ScaledObject`, `ScaledJob`, `TriggerAuthentication`), 80 documented scalers (counted at v2.21), HPA behavior tu

This skill covers running, configuring, deploying, and integrating with **Keycloak**, the open-source identity & access management server. It targets the modern **Quarkus-based** distribution (24.x → 26.8.x); for instances still in the field on 16.x–23.x — including the WildFly **"-legacy"** b

Target: operating LiteLLM proxy with 2+ replicas (typically the in-repo Helm chart) where Redis or Valkey is supposed to make rate limits, spend/budget enforcement, cooldowns, and locks fleet-wide. Grounded in source at `4d543245` (v1.95.0-dev, 2026-07-29; latest stable **at that time** v1.94.0) plu

Target: operators scripting a LiteLLM proxy's control plane — key/team/user/org lifecycle, model management, budgets, spend — instead of clicking the Admin UI. Grounded in source @ `4d543245` (v1.95.0-dev, 2026-07-29) + docs corpus + issue-tracker sweep of 2026-07-30. **Latest stable is v1.106.5

`POST /v1/messages` is served natively by vLLM, SGLang, llama.cpp, Ollama, mistral.rs, Llama Stack/OGX and Lemonade, and adapted by LiteLLM, Bifrost and Superagent Gateway; Claude Code is the driving client. Sibling skills: **`chat-completions-api`** (the surface this one is translated to or from) a

Plan, sequence, and de-risk a **community-edition Grafana Mimir** upgrade deployed via the `mimir-distributed` Helm chart, in an **air-gapped** environment. The hard part is not `helm upgrade`. It is that chart 6.0 changes the **write-path architecture**, the **proxy's DNS name**, and the **admissio

Opinionated greenfield recipe for **NVIDIA datacenter GPUs on Ubuntu 24.04 LTS** — get from a clean OS install to a healthy host where `nvidia-smi` reports all GPUs, `nvidia-fabricmanager` is `active (running)`, and the gpu-operator `cuda-validator` pod passes. Air-gap is the primary case; connect

Target: operators who script Open WebUI (create users, reconcile model catalogs, drive RAG pipelines, export configs) instead of clicking the admin UI. Grounded in v0.11.0 source (2026-07-27), with mounts, auth, floors and new routes re-checked at v0.13.0 (2026-11-04); the API is officially "experim

Choose and wire open-weights embedding + reranker models into Open WebUI's RAG pipeline, across TEI, vLLM, OpenRouter or LiteLLM. Covers model selection on current MTEB data (quality, zero-shot contamination, and a FLOPs-based throughput model, since no leaderboard publishes speed), the query/document prefix strings each model needs and the PREFIX_FIELD_NAME mode switch that silently eats them, the exact wire shapes Open WebUI sends (URL auto-append on embed but NOT rerank; strict Cohere rerank shape), per-backend cliffs (LiteLLM encoding_format=null, TEI 422s, vLLM registry gaps and pooling fallbacks, OpenRouter's misreported rerank pricing), and copy-paste config per model.

Snapshot: OpenShell **0.1.7** (2026-11-03), from the v0.1.7 tag, release notes, GitHub issues filed since 0.1.0, and practitioner write-ups (history in `references/sources.md`). OpenShell releases about weekly and breaks things between minors — **check the installed and latest versions before advi

Run, configure, upgrade and troubleshoot the Zalando postgres-operator (acid.zalan.do `postgresql` CRs, Spilo, Patroni, WAL-G) on Kubernetes. Core knowledge: the v1.x -> v2.x major upgrade and its three changed defaults that bite silently (`kubernetes_use_configmaps` true, spilo-17 -> spilo-18, scram-sha-256 password encryption); the Patroni DCS living in deprecated v1 Endpoints and the split-brain hazard in switching to ConfigMaps; Helm never updating the CRDs; exactly which spec changes trigger a rolling update of Spilo pods; in-place major version upgrades and their success/failure annotations; PDBs, delete protection, password rotation, logical backups, maintenance windows.

Migration reference, verified 2026-07-22. Target: **upstream ≥6.7.0** (6.6.0 fixes the CVE but breaks newline-containing passwords, #2254 — 6.7.0 has the corrected fix). Version matrix authority: `k8s-components-checker` `references/compat/rancher-logging.md`. Day-2 configuration of the migrated

Reference for the OpenAI Responses API (/v1/responses), OpenResponses open standard, and Codex CLI. Covers the request/response schema, previous_response_id, Conversations API, server-side compaction, WebSocket transport, hosted Shell tool, Skills, tool_search, MCP connectors, prompt caching, phase field, 59 typed streaming events, 10-backend support matrix (vLLM, llama.cpp, mistral.rs, Ollama, LiteLLM, SGLang, Llama Stack, TensorRT-LLM, Bifrost, Lemonade), and Chat Completions translation with 17 gotchas.

Operate, configure and upgrade Rook-managed Ceph on Kubernetes (CephCluster CR; rook-ceph, rook-ceph-cluster and ceph-csi-drivers charts; ceph-csi-operator). Core knowledge: the Rook -> Ceph -> key-rotation upgrade order and its version gates; the CVE-2025-30156 fix (new AES256K CephX key type, daemon key rotation, the six AUTH_INSECURE_* health codes, kernel 7.0 for CSI keys); the Rook v1.20 move of CSI to ceph-csi-operator and the new ceph-csi-drivers chart; disabling the rook mgr module before Tentacle; Helm monitoring RBAC.

Target audience: operators running **vLLM and/or SGLang inference on Kubernetes**, fronting workers with a router that does cache-aware load-balancing, optional prefill-decode disaggregation, and dynamic worker registration. Especially: hosting **multiple replicas of the same model** behind one addr

snmp_exporter is a **multi-target proxy**: one central instance walks SNMP on many devices; Prometheus tells it what to scrape per request via URL params (`/snmp?target=<ip>&module=<m1,m2>&auth=<name>`). Everything below assumes v0.30.x. The generator and exporter versions **must match** — a confi