Skip to content

prompt-catalog.md

Latest commit

 

History

History
91 lines (68 loc) · 3.86 KB

prompt-catalog.md

File metadata and controls

91 lines (68 loc) · 3.86 KB
Raw
kind doc
output docs/ai-prompt-catalog.md
description Index of the reusable AI prompts and how to use them in each tool

AI prompt catalog

This repository ships reusable prompts for generating NetApp storage automation: one for planning an API sequence, one per tool for generating code, one that generates all four at once, and one for reviewing the result before a PR.

Each prompt is authored once in ai/ and generated into every format the supported assistants read, so the same prompt is available whichever editor you use.

Where the prompts show up

Tool Reads How to invoke
GitHub Copilot (VS Code, github.com) .github/prompts/*.prompt.md Type / in Copilot Chat
Cursor .cursor/commands/*.md Type / in the chat panel
Any other assistant the prompt files themselves Paste the file contents

Commands are named <product>-<task>, so typing /ontap- lists everything scoped to ONTAP.

Available prompts

{{PROMPT_FILES}}

For an assistant with no slash-command support - ChatGPT, Gemini, Claude, and the like - open the prompt file in the last column and paste its contents into the chat, then replace {task description} with your task. The file is the prompt; there is no separate copy to keep in sync.

Using one

  1. Open the chat panel, type /, and pick the command.
  2. Replace {task description} with your storage task.
  3. Work through the prompt's steps. Most stop and ask you to confirm the API sequence before they write any code - that checkpoint is the point.

A typical run through a new use case:

1.  /ontap-plan-api-sequence   →  Design and validate the REST API sequence
2.  /ontap-generate-workflow   →  Generate Python + Ansible + Terraform + Go
    (or /ontap-generate-python, -ansible, -terraform, -go individually)
3.  /review-contribution       →  Check conventions, CI compliance, README updates

Conventions attach on their own

Product conventions are not prompts you invoke. They are attached automatically when you edit files in that product's directories - through .github/instructions/ for Copilot and .cursor/rules/ for Cursor - so a file under python/ontap/ picks up the ONTAP rules without you asking.

Repo-wide context works the same way through AGENTS.md, which Cursor and most other agents read, and .github/copilot-instructions.md for Copilot.

AI output is a draft

Every prompt produces a starting point, not a merge. Run make ci, validate against a real cluster, and record the result in the PR's Test Report - the same bar as any hand-written change.

Editing the prompts

Everything under .github/prompts/, .github/instructions/, .cursor/, plus AGENTS.md, .github/copilot-instructions.md, and this page, is generated. Edit the source in ai/ and run make ai-assets; CI fails if the two drift apart.

Task description cheat sheet

Prompts expect a one-line task description. Examples that work well:

Category Example task description
NFS Create an NFS volume with a dedicated export policy and client-match rule
CIFS Create a CIFS share on an existing volume with read/write ACL for a domain group
iSCSI Provision an iSCSI LUN with an igroup and map it to a specific initiator
Snapshot Create an on-demand snapshot of a volume and list all snapshots
Cluster Retrieve cluster health: node status, aggregate usage, and version info
SVM Create a new SVM with NFS and CIFS protocols enabled
Volume ops Clone a FlexVol volume from an existing snapshot
SnapMirror Set up SnapMirror replication between two SVMs on different clusters
QoS Create a QoS policy group and assign it to an existing volume
Resize Resize an existing volume and verify the new capacity