ariadne-kubernetes-graph

Ariadne Kubernetes Graph

Overview

Use this skill when the task is about Kubernetes state that is already available through Ariadne MCP.

The goal is to answer cluster questions with the cheapest reliable sequence of Ariadne tool calls, while avoiding unnecessary schema and health fetches.

When To Use

Use this skill for:

• live cluster questions answered through Ariadne MCP
• snapshot-backed cluster questions answered through Ariadne MCP
• translating operational questions into graph_query
• debugging failed or ambiguous Ariadne queries
• deciding when graph_schema or graph_health is actually needed

Do not use this skill for:

• codebase-only tasks
• generic Kubernetes explanations that do not depend on Ariadne data
• tasks where the user already provided the exact Cypher to run and no Ariadne-specific reasoning is needed

Default Workflow

1. Decide whether the question is really about Ariadne-backed cluster state.
2. Reuse prior successful Ariadne results from the same conversation when still relevant.
3. Call graph_query directly if the labels, edges, and property paths are already known.
4. Call graph_schema only when query shape is uncertain, when working in an unfamiliar part of the graph, or after a schema or semantic failure.
5. Call graph_health only when freshness, readiness, coverage, or sync lag materially affects the answer.
6. Keep queries narrow and use Cypher LIMIT for exploratory reads.
7. Present results in a concise, operator-friendly format instead of dumping raw tool output.

Response Style

When answering through this skill:

• start with the direct answer or next action, not with tool narration
• prefer short headings only when they add clarity
• use flat bullets for multi-part findings
• summarize the result before listing rows or examples
• use fenced cypher blocks only when the query is useful to show or debug
• avoid dumping raw graph_health JSON unless the user explicitly asked for it
• mention freshness, degraded coverage, or sync lag only when it affects trust in the answer
• if a query was repaired, explain the fix in one short sentence
• if the result is inconclusive, say what is missing and what narrower follow-up query would resolve it