/analyze — Evidence-Based Diagnosis

The /analyze skill examines a captured session using a voice-first workflow. It extracts user intent from voice transcripts, then correlates findings with function traces and screenshots for evidence-based diagnosis.

Usage

/analyze

Analyzes the most recent session. Runs automatically after /check, but you can also invoke it independently on any captured session.

How It Works

Step 1: Intent Extraction

ReadyCheck transcribes your voice recording and identifies issues — moments where you expressed confusion, frustration, or described unexpected behavior.

Each issue gets:

A severity level (CRITICAL, HIGH, MEDIUM, LOW)
A time window in the session
Keywords and direct quotes from your narration

Step 2: Issue Selection

For non-critical issues, ReadyCheck asks which ones to investigate. Critical and high-severity issues are always analyzed.

Step 3: Multi-Track Analysis

For each selected issue, ReadyCheck:

Queries function traces in the time window — looking for errors, exceptions, unexpected sequences
Captures screenshots at key moments — visual evidence of what was on screen
Correlates all three tracks — voice intent + visual evidence + code behavior

Step 4: Findings Report

Results are presented in a structured summary:

Issue	Severity	Primary Cause	Likelihood
ISS-001	HIGH	Missing null check in save handler	High

Each issue includes:

Primary hypothesis with confidence score
Evidence from traces and screenshots
Suggested code areas to investigate

Step 5: Fix Planning

After reviewing findings, ReadyCheck creates a fix plan with:

Verification steps for each hypothesis
Specific code changes
Regression test cases

No Voice Recording?

If the session was captured without voice, ReadyCheck switches to trace-first analysis — scanning all trace events for anomalies like exceptions, long gaps (potential hangs), and unexpected function sequences.

No Screen Recording?

Analysis continues with traces and voice. Findings note that visual correlation is unavailable.