Deepen it-operations: virtualization, cloud, telephony, tiering, multi-site

[cbrock84]

Corporate IT owned five things the catalog had no answer for. Three become new skills; two are gaps inside skills that already existed. it-operations goes from 8 skills to 11, and the catalog from 143 to 146.

New skills

virtualization-operations — the hypervisor layer beneath the servers. systems-administration covered servers generically and nothing covered the layer they run on: consolidation ratio as a decision rather than a discovered fact, N-1 capacity, VM sprawl as a lifecycle problem, snapshot discipline (old snapshots fill the datastore and take down every VM sharing it), live migration as a maintenance capability rather than a resilience one, and licensing that counts physical cores rather than instances.

cloud-administration — the cloud the company runs on, which is a different discipline from the one it sells. technology:cloud-infrastructure keeps the product side (IaC, environments, region failure); this owns tenant and subscription structure, the SaaS estate nobody is managing, identity as the actual perimeter, spend that arrives as a surprise, and the narrowness of shared responsibility. The split follows the CIO/CTO boundary already established in D27 — the CIO runs the technology the company works on, the CTO the technology it sells. Both skills cross-reference each other so the boundary stays legible.

telephony-and-conferencing — previously absent entirely, despite being IT-owned in essentially every mid-market organization. Voice as a real-time application on a network built for something else, emergency calling as a legal obligation that hybrid work quietly breaks, number porting and why cancelling early strands numbers, conference rooms failing on the last two feet, and recording creating discoverable records.

Enrichments

service-desk — tier design. The skill covered intake, triage, prioritization and escalation, but not how the tiers themselves get drawn. Added the useful definition (what a tier may resolve without escalating, not seniority), both failure modes with their tells, the argument for drawing the boundary at access rather than difficulty, and why each tier needs its own definition of done.

network-administration — multi-site. A second location introduces failure modes the first never had. Added the three services that hurt most on a link drop (authentication, name resolution, print), partial link failure being harder than complete failure, SD-WAN as a real gain and a real concentration of risk, and small sites accumulating exceptions.

Descriptions kept in agreement

The department description understated its remit in three places that have to agree — plugin.json, marketplace.json, and the agent charter in .claude/agents/. All three move in this commit. This is the same divergence class flagged on #18, so it seemed worth not reproducing.

Verification

• All nine checks in scripts/check-all.sh pass. The US English check caught a "Cancelling" before it landed.
• Generated artifacts regenerated from the tree walk: README, org chart (md, html, both PNGs), social card.
• Every cross-reference resolves — 94 checked.

Provenance

These were written from scratch. The gaps were identified by reading real IT job descriptions for what the roles are accountable for; none of that text appears here, and I verified it: no source organization name, no individual's name, and no industry-specific product names from the source material appear anywhere in the repository.

Worth stating plainly for the record: the source corpus was three usable cross-industry IT roles from a single company. That was enough to identify what the catalog was missing — every one of those roles owned virtualization, cloud and telephony, and every one defined itself by support tier. It is not enough to validate the content, which is written from general practice rather than from the corpus. Read it as you would any other assertion in this repo.

---

Generated by Claude Code