telemetry

The telemetry skill analyzes per-agent execution telemetry from CLAUDE CLI JSONL logs to measure token usage, estimate API costs using Claude pricing, and compute performance trends across all sr-* agents in Markdown or JSON format.

Yes, you can generate a performance dashboard from JSONL session logs. The skill computes per-agent metrics like success rate and average duration, rendering trend-aware dashboards in Markdown or JSON with optional persistent snapshots for future comparisons.

The best way to compare token usage and throughput across multiple agents is to aggregate per-agent execution telemetry. This skill unifies metrics like run count, token usage, and estimated costs to compare agents and identify optimization opportunities.

The telemetry skill requires no external dependencies, but it needs access to CLAUDE CLI JSONL session logs and agent-memory metadata to parse logs, aggregate metrics, and estimate costs effectively.

Yes, you can filter agent telemetry data by specific time periods or individual agents. The skill supports optional flags for period and agent selection, allowing you to narrow down the analysis and focus on specific performance trends.

API cost estimates vary across Claude agents because the skill calculates costs using Claude pricing by model. It provides a per-model breakdown, meaning different models used by different agents will result in varied cost estimations for token usage.