`sparkline`

Turns a short run of numbers into a tiny inline SVG that shows the trend at a glance.

Turn a short series of numbers — a benchmark trend, a metrics readout, a latency/throughput/error-rate history — into a tiny inline trend chart as a self-contained SVG. Reach for this whenever a task hands over a run of numbers and the honest artifact is a single small chart showing the shape of the trend rather than a wall of digits (a benchmark note, a metrics summary, a README stat, a status line). Never a slash command; use it implicitly when the intent calls for one.

What it does

sparkline takes a short series of numbers — --data "12,18,9,22,15,27" — and draws its trend as one tiny inline chart: a smooth polyline over a faint gradient fill, scaled to its own min/max, with the last point marked by a dot. The output is a self-contained SVG (scalable, self-sizing, verified as valid XML), a light teal stroke on a subtle dark panel that reads on the tool pages; --color, --label, --width, --height, --no-baseline and --bare tune it.

It is a tool, not a command: you never type it. The crew reach for it on their own when a task hands over a run of numbers — a benchmark trend, a metrics readout, a latency/throughput/error-rate history — and the honest artifact is the shape of the curve rather than a wall of digits. It has no dependencies beyond python3 and the standard library, so a run is deterministic and offline.

Examples

A few things sparkline can produce, each from a small spec of its own.

A declining sparkline in coral: p95 latency falling from 182ms to 88ms across a release, the last point marked with a dot. — A latency history that improves across a release — p95 falling from 182ms to 88ms.

A rising sparkline in teal: throughput climbing from 420 to 845 requests per second, the last point marked with a dot. — Throughput climbing as autoscaling kicks in — 420 to 845 req/s.

A sparkline in gold with a single sharp spike: error rate jumping to 2.1% during an incident, then recovering to 0.2%. — An error-rate spike during an incident, then recovery — a lone peak at 2.1%.

How the crew reach for it

Tools are opt-in. Add it at install time with shipmates install --harness <name> --with-tools sparkline, or omit the flag and pick it from the interactive list. Once installed it sits alongside the crew as a capability they invoke implicitly; there is no slash command to type.

The renderer and its instructions are bundled together, so wherever the tool lands the agent has both the sparkline.py script and the CLI it needs to drive it: python3 sparkline.py --data "…" --out spark.svg.

Reference

Name: sparkline
Description: Turn a short series of numbers — a benchmark trend, a metrics readout, a latency/throughput/error-rate history — into a tiny inline trend chart as a self-contained SVG. Reach for this whenever a task hands over a run of numbers and the honest artifact is a single small chart showing the shape of the trend rather than a wall of digits (a benchmark note, a metrics summary, a README stat, a status line). Never a slash command; use it implicitly when the intent calls for one.
Bundled files: sparkline.py
Install: shipmates install --harness <name> --with-tools sparkline

Where this lives

This page is generated from toolbox/sparkline/tool.md. Opt in at install with --with-tools sparkline and the installer copies the tool — instructions and bundled files — into the harness's skills tree.

View toolbox/sparkline/tool.md on GitHub

Other tools

All tools