AI Coding Agents Observability And Evals

Use this skill to design or review the feedback loop around a coding-agent runtime: traces, replayable transcripts, eval packs, regression gates, tool-call grading, latency and cost accounting, and production failure triage.

This skill covers how you operate a coding-agent product after the core runtime exists. It does not replace the runtime skills themselves.

ASCII Flow