zicato-read-telemetry

Trace zicato evaluation runs by correlating events.jsonl telemetry with loss.json profiles.

4|2|Updated May 14, 2026
One-click install
npx skills add https://github.com/pedapudi/zicato --skill zicato-read-telemetry
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: zicato-read-telemetry
Source: https://github.com/pedapudi/zicato/tree/main/skills/zicato-read-telemetry
Command: npx skills add https://github.com/pedapudi/zicato --skill zicato-read-telemetry

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps you understand what happened during zicato evaluations by tracing canonical run telemetry, reduced loss profiles, structured invocation logs, and harmonograf sessions without relying on stale derived indexes.

Core Features & Use Cases

  • Run Tracing: Follow a specific epoch, generation, and board entry through its events.jsonl and loss.json files.
  • Loss Interpretation: Examine drift counts, judge attribution, plan revisions, pass/fail results, metrics, runtime data, and abort conditions.
  • Session Correlation: Distinguish zicato's meta-loop session from per-board-run sessions and build harmonograf deep links using the recorded session identifier.
  • Operational Guardrails: Use canonical telemetry files, recognize dialect-specific behavior, inspect structured logs, and avoid starting live evolution runs merely to generate telemetry.

Quick Start

Use the zicato telemetry skill to trace a specific board run, interpret its loss profile, and open the corresponding live harmonograf session.

Frequently Asked Questions about zicato-read-telemetry

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I trace multi-agent evaluation runs using telemetry events?

To trace multi-agent evaluation runs, you correlate canonical events.jsonl telemetry with reduced loss.json profiles and harmonograf sessions. This allows you to track specific epochs, generations, and board entries without relying on stale derived indexes.

What is the best way to diagnose drift and loss patterns in multi-agent systems?

Diagnosing drift and loss patterns involves examining reduced loss profiles to interpret drift counts, judge attribution, plan revisions, pass/fail results, metrics, runtime data, and abort conditions recorded during the evaluation run.

How do I distinguish meta-loop activity from per-board-run sessions in observability data?

You can distinguish meta-loop activity from per-board-run sessions by correlating canonical run telemetry with structured invocation logs and harmonograf sessions, using recorded session identifiers to separate the overarching loop from individual board runs.

How can I create deep links to live harmonograf dashboard sessions?

You create harmonograf deep links by extracting the recorded session identifier from the canonical telemetry files and using it to construct a direct URL to the corresponding live dashboard session for the specific evaluation run.

Can I generate telemetry data by starting a live evaluation run?

No, you should avoid starting live evolution runs merely to generate telemetry. Instead, you must read existing zicato workspace artifacts, interpret telemetry dialects, and inspect structured logs to analyze past runs across epochs and generations.