TraceAI CLI Diagnostics and Reports

Problem Addressed

After agent tool calls, MCP events, and SDK activity, engineers often need a stable entry point to confirm whether traceai is available, whether storage is healthy, and whether failures point to parameter design rather than code bugs. This skill organizes installation, validation, metrics, and reporting into actionable steps.

How It Works

It treats traceai as the default command and favors local builds with SQLite storage, using traceai.db as the default database name. A typical path is to confirm the binary and version injection with version, check storage and service health with health, inspect runtime metrics with metrics, then analyze tool heat, error rates, behavior profiles, and agent usage through report and export. Reports commonly include Tool Heatmap, Error Rate Ranking, Behavior Profile, Failure Reasons, Agent Usage, and Trend; exports are CSV or JSON for spreadsheet review or further analysis.

Boundaries

It is suitable for command execution, release validation, configuration checks, database path, permission, and log troubleshooting. It is not for product roadmap discussion, feature-value debate, or tasks that do not require running commands. If health fails, inspect storage path and permissions instead of assuming missing data; empty export output does not mean command failure; do not seed demo data by default.