Inspect local coding-agent traces and share minimized reports with TraceCrate
Use TraceCrate to load Claude Code, Codex, OTLP, or native trace files locally, compare agent runs, inspect tool timelines, and export privacy-conscious structure-only reports.
npx skills add agentskillexchange/skills --skill inspect-local-coding-agent-traces-and-share-minimized-reports-with-tracecrate
Use TraceCrate when an operator needs to review what a coding agent actually did after a run without uploading private transcripts to an observability service. TraceCrate reads selected local trace files, builds a searchable event timeline, compares runs, surfaces reported metrics and diagnostics, and exports minimized reports with optional timing and token removal.
What this skill actually does
The repeatable workflow is concrete: open the hosted demo or clone the repository, run the local Vite app, choose authorized Claude Code JSONL, Codex rollout JSONL, OTLP JSON, or TraceCrate native files, inspect the sequence and tool events, compare runs, then export a structure-only report for review. Invoke it instead of using a normal cloud tracing product when the source of truth is existing local agent logs and privacy boundaries matter. The scope boundary is post-run trace inspection and deliberate report export, not running agents, instrumenting applications, collecting telemetry, or serving as a generic observability platform.
Inputs and prerequisites: TraceCrate, Node.js 22.12 or newer for local use, npm, and authorized local Claude Code, Codex, OTLP JSON, or TraceCrate native trace files..
Setup notes: For local use, run git clone https://github.com/FankChen/tracecrate.git , cd tracecrate , npm install , and start the Vite app with the project command documented in the README. Open the printed local address, then choose or drop supported trace files you are authorized to inspect.
Source and verification boundary: use https://fankchen.github.io/tracecrate/ as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.
Framework fit: publish this as a Multi-Framework workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.