ai-app-instrumentation
Installation
SKILL.md
Coralogix AI Center — instrument, run, verify
Takes a GenAI application at any maturity — from no OpenTelemetry to a production
collector — in any language, provider, or framework. Adds gen_ai.* OpenTelemetry
spans, gets the app running, and verifies the spans render correctly in AI Center.
Out of scope: whole-service APM beyond the AI call paths; the Code Agents screen (Claude Code/Cursor CLI telemetry); zero-code eBPF capture.
When to Use This Skill
| Use case | What to do |
|---|---|
Add gen_ai.* spans to an app that has none, or has OTel but no GenAI spans |
Full flow: Phase 1 Assess → Phase 2 Instrument → Phase 3 Run → Phase 4 Verify |
| An app already emits GenAI spans; "verify", "audit", "pull the spans", app missing from AI Center, no cost or prompts | Verify-only mode (Phase 4 §1b) — no repo changes, no questions beyond scope |
| Choose the library or gateway that produces the spans for a stack | Research ladder in Phase 2, then the Stacks table |
| Spans flow but render wrong (empty conversations, zero cost, one provider for every model) | Pitfalls table, then the gate whose criteria failed in Phase 4 |
| Claude Code / Cursor / Copilot CLI usage telemetry (Code Agents screen) | Not in scope — that is the CLIs' own telemetry, not gen_ai.* spans |
| Whole-service APM, collector deployment | Not in scope — use opentelemetry-collector / opentelemetry-instrumentation |