AI Prompt Library — Free Prompt Templates

A prompt library is only as good as the structure behind each template. Every entry in this library follows the same four-section pattern — Role, Task, Constraints, Output format — that Anthropic, OpenAI, and Google's published prompt engineering guides all converge on. The library doesn't call any AI; it's a static dataset that lives in your browser, with a small in-page tool to fill in placeholders and copy the result.

1. What each template provides

Every prompt ships with five pieces: a title describing the outcome, a one-line summary, the body (with {{placeholders}} for the variables you supply), a when-to-use note that tells you when the template is the right choice, and a source citation — which of the published prompt engineering guides most directly shaped its structure.

2. Categories

48 templates are organised into six categories. Each card carries a category badge and the count next to the filter chip:

• Writing (8 templates) — Blog drafts, emails, summaries, headlines, rewrites.
• Coding (8 templates) — Code review, debugging, refactors, tests, regex, SQL.
• Analysis (8 templates) — SWOT, pros/cons, decisions, root-cause, comparisons.
• Learning (8 templates) — Study plans, quizzes, Feynman, concept maps, exam prep.
• Marketing (8 templates) — Ad copy, landing pages, SEO briefs, personas, product copy.
• Productivity (8 templates) — Meeting prep, status updates, decision docs, OKR drafts.

3. Placeholders

Anything wrapped in {{double_braces}} is a variable you fill in. Open any template card to see the side form: each placeholder has a one-line hint and a realistic example. Click Use examples to auto-fill every field, or fill some and leave the rest. Unfilled placeholders remain as {{name}} in the output — useful if you want the AI to interpret them or you plan to edit before sending.

4. Source-cited methodology

Every template is labelled with the published guide whose structure most directly shaped it. The references are linked at the bottom of this page and on each template's detail panel. The four-section structure (Role / Task / Constraints / Output) comes from Anthropic's prompt engineering overview; the banned-phrases lists in some writing prompts come from OpenAI's style guidance; the comparison-and-reasoning prompts cite the Wei et al. 2022 chain-of-thought paper.

5. Integrity check (the cross-verifier)

The data module exports verifyLibraryIntegrity() — a deterministic self-check that runs through every template and asserts: each one has a non-empty title, body, summary, and when-to-use; every placeholder that appears in the body is declared in the template's metadata (and vice versa); and renderPrompt(body, {}) returns the body unchanged. Running it on this library returns an empty list of issues — that's the cross-check equivalent of the IRD formula reconciliation used in our tax calculator.

Everything is deterministic. The same template plus the same variable values always produces the same copyable prompt — no randomness, no LLM call, no rate limit.

• Anthropic — Prompt engineering overview (Claude API docs)
• Anthropic — Use XML tags to structure prompts
• OpenAI — Prompt engineering guide (developer platform)
• OpenAI — Six strategies for getting better results
• Google — Gemini API prompting strategies
• Wei et al., 2022 — Chain-of-Thought prompting (arXiv:2201.11903)
• Brown et al., 2020 — Language Models are Few-Shot Learners (GPT-3, arXiv:2005.14165)

Templates and their cited sources were last cross-checked on 2026-05-12. The library is reviewed quarterly and whenever a referenced provider publishes a substantive update to its prompt engineering guidance.