What does good API documentation include?

Each endpoint needs its method and path, a description, parameters with types and whether they're required, an example request, an example response, and the possible error codes. The prompt enforces that structure so nothing is missed.

How detailed should a README be?

A strong README covers what the project does, how to install it, a minimal usage example, configuration, and how to contribute. The prompt scales detail to your reader level so a beginner gets more setup hand-holding than an expert.

What makes a runbook actually useful in an incident?

A runbook must be scannable under stress — clear symptoms, numbered steps, exact commands, expected output at each step, and an escalation path. The prompt produces copy-pasteable commands and a rollback procedure rather than prose.

Is my code sent anywhere?

No. The tool only generates a text prompt in your browser. You paste your code into your own LLM; nothing is uploaded or stored by this tool.

Technical Documentation Prompt Pack

Technical documentation prompt pack

Engineers write great code and mediocre docs because docs are a different craft — they reward structure, completeness, and the discipline to write for someone who knows less than you do. This pack gives you battle-tested prompts for the four documents teams write most: API reference, README, runbook, and changelog. Each prompt encodes the conventions reviewers expect so the AI’s first draft is close to shippable.

How it works

You pick the document type and set the technology or language plus the reader’s experience level. The builder loads a structure proven for that type — for an API endpoint, that means method, path, parameters with types, example request and response, and error codes; for a runbook, symptoms, numbered remediation steps, exact commands, and a rollback path. It then writes a prompt that tells the model to follow that structure, match your stack’s idioms, and pitch the explanation at the right level. You paste your code, logs, or notes and the model fills the template accurately.

Tips and examples

Feed it real signatures. Paste the actual function or endpoint definition so the model documents what exists, not what it guesses.
Set the reader level honestly. “Beginner” gets install steps and concept explanations; “expert” gets terse reference. Mismatch is the top docs complaint.
Runbooks: demand exact commands. The runbook prompt asks for copy-pasteable commands and expected output, which is what you need at 3am, not paragraphs.
Changelogs: group by type. The changelog prompt sorts entries into Added, Changed, Fixed, and Removed in Keep-a-Changelog style so releases read cleanly.