Tools

• collect_cluster_status: modes snapshot (default) | windowed
  • Collection only — diagnosis and recommendations are produced by this skill.
  • For charts, delegate to the visualization skill; never emit chart specs directly.

Workflow

1. Always call collect_cluster_status before any health assessment.
2. Use status_analysis_mode="windowed" for bounded-time questions (e.g., "past 3 hours"); keep the same window in follow-up calls.
3. Base all findings solely on collect_cluster_status output.

Severity Thresholds

Output Format

1. ### Summary — print this heading, then:

   Status	Nodes with Issues	Checks Run	Timestamp
   🟢 OK / 🟠 WARNING / 🔴 CRITICAL	N	comma-separated category names (e.g., parts, replication, disk)	ISO8601

2. ### Findings by Category — print this heading, then a markdown table with one row per category returned by the tool, in stable order:

   Category	Status	Key Metrics	Top Outlier / Scope	Notes

   Rules:

   • Status: emoji + text (e.g., 🟠 WARNING), never emoji-only.
   • Key Metrics: 1–2 values, single-line, semicolon-separated (e.g., max_parts_per_partition=533 (>500)).
   • Notes: remaining metrics as compact key=value items, single-line.
   • Wrap identifiers in backticks (e.g., `db.table`).
   • No outlier → set Top Outlier / Scope to -.
   • No multi-line content inside cells.

3. Recommendations — max 3 items: title + reason + SQL/command if applicable.

Rules

• Never give a health opinion without calling collect_cluster_status first.
• Never assume schema or table names; use only tool output.
• Never write custom health-check SQL; the tool is the source of truth.