System Prompting

Two LLM-readable endpoints

Every public grid ships two plain-text surfaces designed to be pasted directly into a system prompt or fetched at agent startup. No API key required.

# Plain text — 40 sample rows, column list, grid metadata
curl https://instadash.io/<handle>/<slug>/llms.txt
 
# Markdown table — same data, human-readable, attribution footer
curl https://instadash.io/<handle>/<slug>/llms.md

What llms.txt contains

grid: raihan/sales-q2
title: Q2 2026 Revenue
description: Consolidated revenue across all regions
rows: 12408  cols: 8
columns: id, region, product, amount, currency, date, rep, status

--- sample rows (40 of 12408) ---
{"id":1,"region":"APAC","product":"Pro","amount":4200,"currency":"USD",…}
{"id":2,"region":"EMEA","product":"Builder","amount":1100,"currency":"EUR",…}
…

What llms.md contains

# raihan/sales-q2 · Q2 2026 Revenue

| id | region | product | amount | currency | date       | rep    | status |
|----|--------|---------|--------|----------|------------|--------|--------|
| 1  | APAC   | Pro     | 4200   | USD      | 2026-04-01 | Alice  | closed |
| 2  | EMEA   | Builder | 1100   | EUR      | 2026-04-02 | Bob    | open   |
…

---
attribution: { "v": 42, "source": "cli", "timestamp": "2026-05-07T09:14:22Z", "sha256": "e3b0c44…" }

Paste into a system prompt

Fetch the /llms.md output and paste it into your system prompt before your first user message. The model will use the sample rows to ground its answers and cite the version for traceability.

You are an analyst with access to the following dataset:

<dataset>
{paste llms.md output here}
</dataset>

Answer questions about this data. Always cite the version number when referencing figures.

Fetch at agent startup

For agents that run continuously, fetch the /llms.txt surface at startup rather than hardcoding it — so the agent always sees the latest version.

const res = await fetch('https://instadash.io/raihan/sales-q2/llms.txt')
const context = await res.text()
// Include `context` in your system prompt