What problem does it solve?
Standing up a DOCA telemetry collector involves four easily-confused surfaces — the CollectX collection mechanism, the doca-telemetry reader library, the doca-telemetry-exporter publisher library, and the productized DTS container — and misconfiguring any stage silently drops metrics. This Skill guides agents through deploying, wiring, and debugging a CollectX-based collector on a host or BlueField without inventing provider names, schema fields, flags, or config paths.
Core Features & Use Cases
- Four-surface decomposition: Separates the clx collection mechanism from the reader library, publisher library, and productized DTS container before any config guidance, routing out-of-scope questions to the correct owner.
- Pipeline and exporter workflows: Step-by-step verbs (configure, run, test, debug) covering providers/counters, schema, collector daemon, and export backends (Prometheus pull, Fluent Bit push, NetFlow, file/IPC).
- Layered error taxonomy and safety policy: Six-layer diagnosis ladder (collector won't start through transport failure), gate-before-commit provider probing, and end-to-end smoke testing before declaring deployment.
- Use Case: An operator's collector daemon runs but their Prometheus consumer shows nothing — the Skill walks the layered ladder (no provider rows, schema mismatch, exporter silent, downstream skew) to isolate the failing layer instead of guessing.
Quick Start
Ask your agent to help you deploy a CollectX-based DOCA telemetry collector on your BlueField and ship counters to a Prometheus endpoint.
