Lists multiple specific concrete actions across tiers: node drain procedures, OS patching, hardware changes, configuration updates, maintenance window configuration, patch scheduling, deferral policies, and monitoring during CRL-managed maintenance.

Clearly answers both 'what' (manages planned cluster maintenance with specific actions per tier) and 'when' with an explicit 'Use when planning maintenance, configuring maintenance windows, or preparing applications for maintenance events' clause.

Includes strong natural keywords users would say: 'maintenance', 'maintenance windows', 'node drain', 'OS patching', 'patch scheduling', 'hardware changes', 'deferral policies'. These cover common variations of how users would describe maintenance-related tasks.

Highly distinctive with a clear niche around planned cluster maintenance across specific tiers (Self-Hosted, Advanced/BYOC, Standard, Basic). The tier-specific details and maintenance-focused triggers make it unlikely to conflict with other skills.

The skill is generally efficient but includes some sections that could be tightened — the Standard and Basic tiers have minimal content that could be collapsed into a single section, and the 'When to Use This Skill' list partially duplicates the opening paragraph. The context-gathering tables are useful but somewhat verbose for information Claude could infer from the routing table alone.

Provides fully executable SQL queries for pre-maintenance checks, concrete bash commands for drain/stop/start across platforms, specific API calls for cloud console operations, and clear conditional branching (if platform = X, use Y). Commands are copy-paste ready with appropriate placeholders.

The Self-Hosted workflow has explicit pre-maintenance checks with clear stop conditions ('STOP if any node is not live'), a sequenced drain → stop → maintain → restart → verify flow, and post-restart verification queries. The Advanced/BYOC sections have clear step sequences. The troubleshooting table provides error recovery guidance. Safety considerations explicitly state 'only drain one node at a time' and other critical constraints.

The skill provides a clear overview with context-driven routing to tier-specific sections, references detailed content to separate files (drain-details.md, maintenance-prechecks.md, safety-guide.md), and links to related skills and official documentation. References are one level deep and clearly signaled. The main content stays at the right level of detail while pointing to deeper resources.