newrelic-entity-scout

name	newrelic-entity-scout
description	Find and validate New Relic entities, establish correct account scope, and bootstrap baseline health metrics with New Relic MCP tools. Use when asked to find an entity GUID, confirm entity identity, troubleshoot entity search syntax/scope, or gather first-pass golden metrics before deeper NRQL or incident work.

New Relic Entity Scout

Use this fixed order:

1. Credential and scope preflight
2. Entity discovery (getEntityByName, then searchEntity fallback)
3. GUID validation (getEntityByGuid)
4. Baseline metrics bootstrap (getGoldenMetricsByGuid)

Load references/discovery-flow.md before running discovery.

Required Inputs

• Entity hint (name or guid)
• Expected entity type/domain when known
• Absolute investigation window

If input is missing, set it to unknown and continue.

Procedure

1) Run Credential and Scope Preflight

• Confirm New Relic MCP tools respond (mcp__newrelic__introspection or a minimal entity call).
• Confirm target account scope for the run.
• If credentials or account scope fail, classify as auth-or-scope and stop.

2) Run Entity Discovery

• If a GUID is not provided:

1. Run mcp__newrelic__getEntityByName with the strongest known name hint.
2. If empty or ambiguous, run mcp__newrelic__searchEntity with minimal query terms.
3. If searchEntity returns INVALID_INPUT for the tested pattern (common with hyphenated bare terms), mark that query pattern as tooling-unverified and continue discovery with narrower getEntityByName hints instead of retrying the same failing pattern.

• If a GUID is already provided, skip to Step 3 and mark discovery as not-needed.

3) Validate the GUID

• Run mcp__newrelic__getEntityByGuid on the selected candidate.
• Confirm name/type/domain match expected target.
• If mismatch or not found, return to Step 2 with narrower query terms.

4) Bootstrap Baseline Metrics

• Run mcp__newrelic__getGoldenMetricsByGuid for the validated GUID.
• Record metric availability and obvious null/empty gaps.
• If no metrics are returned, classify as no-baseline-metrics and report next action.

Output Contract

Return:

1. status (resolved, blocked, or partial) and primary failure class if blocked.
2. Evidence table with step, tool, input, result, and verified|inference.
3. Final entity block (guid, name, type, domain).
4. Baseline metric summary from getGoldenMetricsByGuid.
5. One next action only.

Guardrails

• Keep ordering deterministic; do not fetch metrics before GUID validation.
• Separate verified tool outputs from inference.
• Mark any non-executed query/filter pattern as unverified.
• Keep scope to entity discovery and bootstrap only.

Non-Goals

• Deep NRQL rewrite workflows.
• Cross-signal incident correlation narratives.