asciinema-analyzer_skill

Overview

This skill performs semantic analysis on converted asciinema .txt recordings to extract keywords, detect patterns, and locate high-density activity windows. It combines fast curated searches with optional unsupervised extraction and density analysis to produce actionable summaries or machine-readable exports. The workflow is interactive and enforces mandatory checks and decision points to ensure reliable, reproducible results.

How this skill works

The analyzer starts with a preflight check to validate input and discover candidate .txt files. It runs a tiered analysis: ripgrep for fast curated keyword counts, YAKE for optional unsupervised keyword discovery, and an optional TF‑IDF/topic step for deeper modeling. Density analysis windows identify peaks of keyword concentration and the tool supports multiple output formats (summary, detailed, JSON, Markdown).

When to use it

• When you need to find specific commands, errors, or terms inside an asciinema session transcript
• To extract topics or recurring themes from developer or demo recordings
• When you want to auto-discover unexpected keywords or patterns in session history
• To create summaries or machine-readable exports for documentation or analytics
• To locate peak activity or dense segments for review or clipping

Best practices

• Always convert .cast files to .txt before analysis and run the preflight check
• Start with curated keyword searches (ripgrep) and only run YAKE when auto-discovery is needed
• Limit YAKE to smaller files or increase system resources; use ripgrep for very large transcripts
• Select domain keyword sets relevant to the session to reduce noise and speed results
• Use density analysis to prioritize where to review or extract context for reports

Example use cases

• Scan a training session transcript for ML/AI terms (epoch, loss, inference) to produce a topic summary
• Identify trading-specific mentions (backtest, drawdown, sharpe) across multiple recordings for compliance reviews
• Find and extract instances of errors or debugging commands from a developer walkthrough
• Auto-discover surprising terms with YAKE to spot new patterns or tooling references
• Generate a JSON report of keyword counts and top peak windows for downstream automation