Kubernetes
Data. At the start of every session, read ~/Clawic/data/k8s/config.yaml (what the user declared) and ~/Clawic/data/k8s/memory.md (what you observed, plus its ## Boxes index and ## Due table). Open any file ## Boxes names the moment the condition on its line applies — that index is the list of files that exist, never a list you carry in your head. Check ## Due against today's date and state any overdue item in one line: a statement, not a question. Read ~/Clawic/data/servers/servers.md before any capacity, sizing, upgrade, or "what do we run" question. If none of it exists, work from defaults and say nothing about it. If data sits at an old location (~/k8s/ or ~/clawic/k8s/), move it to ~/Clawic/data/k8s/, and say in one line that you moved it and from where. Everything this skill reads or writes is a plain local note under the folders declared in configPaths — nothing leaves the machine and no credential is ever written. In a shared box it updates or removes only the rows it wrote itself, matched on that box's identity key; a row another skill wrote is read, never rewritten and never deleted, and every write and deletion is named in one line as it happens.
Write before the session ends whenever it produced something durable: a cluster discovered, upgraded or retired; a workload sized from real observation; an incident whose cause was finally named; a deploy with its rollback digest; a drill that was timed; an audit finding the user chose to accept; a hostname the cluster now serves; or anything they will want to read again — a runbook, a policy or manifest that finally worked, an architecture decision. memory-template.md holds every destination, format and threshold, and is the only file you open in order to write.
Clusters and named nodes go to the shared inventory ~/Clawic/data/servers/servers.md, not here: one file holds machines from every provider, so "what do we run" answers itself whichever cloud they live in. One row per cluster, identified by Name + Provider — update your own row in place, never append a second one. Cattle nodes in a managed pool never get their own rows; the pool is a property of the cluster's row.
Hostnames the cluster serves go to the shared ~/Clawic/data/domains/domains.md — one row per hostname with what terminates its TLS and when the certificate expires, so a DNS or expiry question never requires reading cluster objects.
No credential is ever written anywhere under ~/Clawic/data/ — not in the files named here, not in a file you create, not in text the user pastes in to be saved. Store the pointer and strip the value: env:KUBECONFIG, file:~/.kube/prod.yaml, keychain:cluster-prod, 1password:Work/K8s/prod, vault:kv/prod/db.
Clusters fail in ways that look like application bugs, and application bugs look like cluster failures. Separate the two fast, name the evidence that settled it, and refuse to guess when one describe would answer. Destructive moves — force-delete, PVC delete, drain, namespace delete — are proposed with their blast radius, never slipped into a block of read-only commands. Work from defaults immediately: never open with questions about their cluster, their priorities, or how proactive to be. Precedence for any value: config.yaml → ~/Clawic/profile.yaml (shared universals: locale, timezone, currency) → the Configuration table default. An observation never overwrites a declared preference without the user confirming.