← All AgentBay walkthroughsAGENTBAY WALKTHROUGH · SYNTHETIC INPUT

Turn a small CSV into a notebook you can rerun

Turn five rows into a notebook the next person can rerun.

Tested with Codex CLI 0.163.0.

START HERE

Install this skill with AgentBay

Read jupyter-notebook SKILL.md ↗ · Pinned publisher instructions

Get the finder once. Install AgentBay for your app, then open a fresh chat.
Install this one skill. Use the button below. The prompt includes this publisher’s exact source and revision. Ask AgentBay to check the instructions, license and prerequisites before installing.
Confirm it is active. Start a fresh agent session if needed. Check that the selected skill appears in Skills, then ask your agent to read its SKILL.md and required resources. In Claude desktop, upload and enable its complete ZIP; a filesystem copy does not activate it.

To follow along, download the starter project, unzip it and open that folder in your agent. Complete the local setup below, then give your agent this task.

Here’s the task to give your agent

Want a few skills? Add them to your selection, review the exact list, and copy instructions for only those choices. Paste the commands into your terminal, then confirm the skills appear in your agent.

What you’re starting with

Load walks.csv and calculate the count, total, mean and totals by route. Explain each step and make the answers come from the data.

Here’s what to do

Ask the agent to use the pinned notebook scaffold, add short explanations and compute the summary from walks.csv. Then restart from a fresh kernel. The recorded Codex run used standard-library calculations, and the independent verifier executed every cell again.

Check the result

The five walks total 10.0 km and average 2.0 km. River contributes 5.5 km, Park 2.5 km and Hill 2.0 km. The independent verifier executes every code cell again and compares summary.json with those exact expected values.

Restart the kernel and run the notebook from top to bottom before sharing it.

Start with the instructions

Read the actual jupyter-notebook SKILL.md ↗

openai/skills · pinned 49f948faa9258a0c61caceaf225e179651397431

Reviewed October 31, 2026.

Good for: Scaffolding an experiment or tutorial notebook
Making a small analysis readable and reproducible from a fresh kernel
Needs: Python 3.12 with the pinned nbformat, nbclient and ipykernel packages
The synthetic CSV and a local Jupyter kernel; no account or external dataset
Scope: Five synthetic observations and descriptive arithmetic
Local notebook executed in a fresh kernel

The input and actual output

Download the synthetic project and unchanged checks (ZIP). It contains no installed third-party skills. Install the pinned upstream skill separately after reviewing its license.

5 walks · 10.0 km total
2.0 km mean per walk
Fresh kernel → exact summary

Inspect the actual solution: walk-analysis.ipynb · summary.json.

Keep the total and the average separate

Ten kilometres sounds like one long walk until you notice it is the total across five rows.

Try this: Distances: 2.5, 1.5, 3.0, 2.0, 1.0 km
Checked result: Total 10.0 km ÷ 5 walks = 2.0 km per walk

Both values are computed in the actual notebook, written to summary.json and checked after fresh-kernel execution. This case shares the notebook’s recorded run and receipt.

Inspect the shared test receipt →

Checks performed

Notebook validates and every code cell executes in a fresh local kernel
Five walks total 10.0 km with a 2.0 km mean
Route totals match River 5.5, Park 2.5 and Hill 2.0 km exactly
Input CSV remains unchanged; analysis imports no third-party data library

Test receipt, installed-file hashes and source binding (JSON)

App test results

Runtime-specific status
Runtime	Result
Codex CLI 0.163.0	passed
Codex desktop	not tested — manual desktop check pending
Claude Code	not tested — no authenticated run recorded
Claude desktop	not tested — manual desktop check pending

Repeat it in a clean project

Unzip the input into a new directory. Follow the setup below; JavaScript examples use Node 24 and Python examples specify their packages. Install only the selected source skill at the pinned revision above. Open a fresh agent session and confirm its skill list contains the installed skill; ask it to read SKILL.md before changing the input.

python3 -m venv .venv
.venv/bin/pip install -r requirements.txt
Ask your agent to scaffold and complete the notebook from walks.csv
.venv/bin/python verify.py

Run the checks, then compare your output with the recorded result.