Prompt – DocsPress

Use docspress/prompt for a reusable AI prompt together with its intended model label, interaction mode, reasoning preference, context, and caption.

When to use it

Choose Prompt when readers should copy and adapt an AI instruction. Use Colorful Code for source code and Terminal Session for commands. Treat prompts as maintained examples: verify their referenced skills, files, and expected outcome.

Context syntax

context is a comma-separated list of up to 12 items. DocsPress classifies each item by prefix:

$skill-name for an installed skill.
@mention for an agent or repository mention.
#image for an image reference.
https://… for a URL.
Any other value as a file or directory.

Invoke a skill as $skill-name. Do not tell an agent to read a SKILL.md path; the installed skill is the portable interface.

Attributes

prompt string required: The complete copyable instruction.

DefaultDocsPress setup review prompt
model string: Display label for the intended model or agent.

DefaultGPT-5
mode enum: Interaction mode shown with the example. Invalid values normalize to chat when rendered.

Defaultcode Valueschat, code, ask, plan
thinking boolean: Shows that deliberate reasoning is expected.

Defaulttrue Valuestrue, false
context string: Comma-separated skill, mention, image, URL, and file references; at most 12 items.

Default$docspress-install, @repository, src/sync.js, docs/ ValuesUp to 12 items
caption string: Formatted explanation below the prompt.

DefaultPrompt example

Creative examples

Find undocumented behavior

Use $generate-docs-from-source to inspect the public exports and identify undocumented behavior.

$generate-docs-from-source @repository src/ test/

Plan an incident-runbook rehearsal

Plan a 30-minute rehearsal for the documentation publishing incident runbook. Include roles, two injected failures, observable success criteria, and a short retrospective template. Do not change production systems.

@operations docs/runbooks/publishing.md test/reconcile.test.js

Review a screenshot for reader friction

Review the attached documentation screenshot from a first-time reader’s perspective. Identify the three moments with the highest cognitive load, explain the evidence visible in the image, and suggest the smallest copy or layout change for each.

#homepage-screenshot @documentation docs/index.md

Published behavior and accessibility

Copy copies the prompt itself. Model, mode, reasoning, and context remain visible metadata, while the containing section receives an AI-prompt label for assistive technology.

Write an explicit deliverable and scope, state what evidence the agent should inspect, and never include a secret in prompt text or context.