Trace Insight — Development skill for Claude Code

Trace Insight

A read-only look at your local agent history.

You have been using coding agents (Claude Code, Codex CLI, Pi) and every one of them has quietly written a transcript of your work to your own disk. This little script reads those transcripts and writes a **Markdown report and a PDF report** about the shape of your work: what you usually do, the tasks you run, the errors you hit, and where your tokens go.

It changes nothing. It reads your history, writes two reports, and stops. No network, no upload, no accounts, no dependencies. Everything stays on your machine. Chrome or Chromium is used only when already installed locally, and its temporary HTML source is deleted after the PDF is created.

---

What you need

• Python 3.8 or newer. Check with:

  python3 --version

  Most Macs and Linux boxes already have it. On Windows, install Python from python.org and use `python` instead of `python3` in the commands below.

• That is the whole list. There is nothing to `pip install`.

Install

There is no install step. Copy the single file `trace_insight.py` anywhere you like, or just run it from the folder you were given.

Install with a coding agent

Give your coding agent this prompt:

Install the Trace Insight skill from https://github.com/casius-connect/trace-insight
as a local skill for this project. Read SKILL.md, then run trace_insight.py. Keep all
agent history local and do not upload, move, or modify it.

For Codex, skills placed in `.agents/skills/trace-insight/` are available to the project. You can also run the script directly from the cloned repository. See the [official Codex skills documentation](https://developers.openai.com/codex/skills/).

Run it

From a terminal, in the folder that contains `trace_insight.py`:

python3 trace_insight.py

That writes these two files to your Desktop:

• trace-insight-report.md: open it in any text editor or Markdown viewer.
• trace-insight-report.pdf: op