ace-study-prompts

The generation rules and self-check checklist that every Ace generator follows. Shipped as a package so the /ace-review Claude Code skill, the hosted ace.study web app, and any third-party generator all produce consistent output.

Part of Ace — open-source infrastructure for AI-generated interactive study content.

Why a separate package

Before v0.1, the generation rules lived only inside .claude/skills/ace-review/SKILL.md. The hosted web app needs the same rules to produce the same quality of output. Copy-pasting invites drift; the eval harness would test one prompt and the hosted app would run another.

This package is the single source of truth. The markdown files are the authoritative content; the JS module re-exports them as strings for any runtime.

Install

bun add ace-study-prompts
# or
npm install ace-study-prompts

Use from JavaScript (Bun, Node 22+, Workers with a text loader)

import { systemPrompt, selfCheck, PROMPTS_VERSION } from "ace-study-prompts";

// Pass to any LLM as the system prompt
const response = await anthropic.messages.create({
  model: "claude-opus-4-7",
  system: systemPrompt,
  messages: [{ role: "user", content: userPdfText }],
});

// Run the self-check after generation
console.log(selfCheck);

Use from Claude Code (or any file-reading agent)

The same markdown files ship in the package so an agent can read them directly:

• packages/ace-study-prompts/src/system-prompt.md — generation rules
• packages/ace-study-prompts/src/self-check.md — pre-flight checklist

The /ace-review Claude Code skill reads these at the start of a run.

Contract

• systemPrompt — the full generation rules as a markdown string. Stable across the v0.x line; content changes trigger an eval-harness run before release.
• selfCheck — the pre-flight invariants as a markdown checklist.
• PROMPTS_VERSION — semver string. Log alongside generations so drift shows up in traces.

License

MIT.