zicato-analyze-epoch

Close completed zicato evaluation epochs and analyze their journals, experiments, telemetry, and retrospective reports.

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

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps you close completed zicato evaluation epochs and understand what happened across their rounds, hypotheses, outcomes, and promotion decisions without rerunning expensive live experiments.

Core Features & Use Cases

  • Epoch Retrospectives: Run or inspect the retrospective analysis for a completed epoch, including its journal and hypothesis-versus-outcome ledger.
  • Report Repair and Rendering: Re-render analysis.md and analysis.html from current on-disk data, with an option to skip auxiliary LLM calls.
  • Telemetry Inspection: Regenerate decision-telemetry insights for a selected epoch or round.
  • Use Case: After an evolution run finishes, use this Skill to close the epoch, review which proposer hypotheses matched observed drift, and repair its reports if earlier generated data is incomplete.

Quick Start

Ask the Skill to close the current zicato epoch, read its journal and analysis, and regenerate the report without using an LLM.

Frequently Asked Questions about zicato-analyze-epoch

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

FAQPage Schema
How do I analyze completed evaluation epochs without rerunning live experiments?

You can analyze completed evaluation epochs by closing them and inspecting their preserved journal, telemetry, and analysis files. This retrospective approach reviews round outcomes and hypothesis-match results without the cost of rerunning live experiments.

How do I repair incomplete historical analysis reports from previous evaluation rounds?

Repair incomplete historical analysis reports by re-rendering analysis.md and analysis.html from current on-disk data. You can regenerate epoch insights and repair reports using preserved artifacts without needing to rerun the original evaluation.

What workspace artifacts are required to perform a retrospective epoch analysis?

Retrospective epoch analysis requires an existing .zicato workspace and preserved epoch artifacts. Specifically, it needs journal.md, analysis.md, analysis.html, and experiment.json files to be present on disk to successfully review outcomes and regenerate reports.

Can I regenerate decision-telemetry insights and skip auxiliary LLM calls during epoch analysis?

Yes, you can regenerate decision-telemetry insights for a selected epoch or round. When re-rendering analysis reports from on-disk data, you can apply an option to skip auxiliary LLM calls for faster report regeneration.

Does epoch retrospective analysis require the zicato CLI to close and review evaluation rounds?

Yes, epoch retrospective analysis requires the zicato CLI to close the current evaluation epoch. The CLI works with the existing .zicato workspace to read journals, validate hypothesis-match results, and regenerate insights.

What is the best way to review which proposer hypotheses matched observed drift after an evolution run?

The best way to review proposer hypotheses against observed drift is to close the completed epoch and inspect its hypothesis-versus-outcome ledger. This retrospective analysis reads the journal and telemetry data to validate promotion decisions.