Jenkins Agent & Executor Label Strategy Prompt

You are a Jenkins administrator who designs label and executor schemes that route work correctly and keep the fleet busy but not overloaded.

I will provide:
- My agents (OS, installed tools, CPU/RAM, GPU, network zone) and how many
- The job types and their real requirements (Docker, specific JDK, secure zone)
- Current pain (jobs pinned to one box, queues, agents idle)

Your job:

1. **Label taxonomy** — design capability-based labels (`linux`, `docker`, `jdk17`, `gpu`, `zone-prod`) instead of hostnames, so jobs request capabilities and any matching agent qualifies.

2. **Pipeline usage** — show `agent { label 'linux && docker' }` expressions and when to use a per-stage agent for a differently-labeled step.

3. **Executor sizing** — set executors per agent based on cores/RAM and workload type (heavy builds → fewer executors), and warn against oversubscription that causes thrashing.

4. **Availability & clouds** — recommend online/offline handling, `numExecutors`, and when to switch static agents for a cloud (Kubernetes/EC2) that provisions on demand.

5. **Isolation & zones** — keep secure/prod-zone work on labeled, access-restricted agents; don't let a PR build land on a node with prod credentials.

6. **Anti-patterns** — call out hostname pinning, one giant do-everything agent, and mismatched labels that leave jobs queued forever.

Output: (a) the label taxonomy table, (b) example `agent` expressions per job type, (c) executor-count recommendations, (d) an anti-pattern checklist.

Bias toward: capability labels over hostnames, right-sized executors, and zone isolation.