Prompt Corpus (CSV / JSON / one Markdown file per prompt)

LLMs

A corpus of 2,130 CC0 public-domain prompts, the same content in three shapes: upstream's CSV, structured JSON, and one Markdown file per prompt; with a category index and a usage guide in Traditional Chinese.

Subscribe & download See how to start

SubscriptionUS$49 per month

Last run2026-11-02

In practice

Readiness5 / 5

What you need

Few changes needed5 / 5

Who it is for

Individuals and small teams putting an LLM into their daily work or their product: they want ready-made starting points for role prompts, and they want them machine-readable, in a shape that can be wired straight into a workflow rather than copied and pasted page by page.

Browse the files

Files in the product repository at the pinned commit. A few house files are left out of the zip.

File overview2,000+ paths

▸prompts1982

0001-ethereum-developer.md
0002-linux-terminal.md
0003-english-translator-and-improver.md
0004-job-interviewer.md
0005-javascript-console.md
0006-excel-sheet.md
0007-english-pronunciation-helper.md
0008-spoken-english-teacher-and-improver.md
0009-travel-guide.md
0010-plagiarism-checker.md
…

▫ATTRIBUTION.md

▫LICENSE

▫LICENSE-99LL.md

▫LICENSE-CC0

▫LICENSE-MIT

▫PRODUCT.md

▫PROMPTS.md

▫QUICKSTART.md

▫README.md

▫SECURITY.md

▫SECURITY.upstream.md

▫delivery-manifest.txt

▫prompts.csv

▫prompts.json

▫使用指南.md

▫分類索引.md

▫Config and hidden files 2

Latest passing run

Clone → Checkout → Bundle inventory → Parse prompt corpus

Pre-purchase Disclosure

What you are buying

A corpus of 2,130 prompts, the same content in three shapes: upstream’s prompts.csv, the prompts.json we generated, and prompts/ with one file per prompt; with 分類索引.md (a category index) and 使用指南.md (a usage guide) in Traditional Chinese.

Built on an open-source prompt corpus project; upstream releases the corpus under CC0 1.0. The upstream website code is not delivered, and the JSON, the per-prompt Markdown and the Traditional Chinese guide are ours. Delivery is a frozen snapshot: one zip file, taken from the product repository at revision 46cc542.

Known limitations

This is a document-type delivery: the health check verifies that the files are complete and can be opened, not what they do when run
The corpus is released into the public domain under CC0 upstream; it is not content exclusive to this product
Of the 2,130 prompts, the category index links to 461; the rest have to be found in prompts.json or prompts/
The longest single prompt is 144,260 characters, so reading prompts.csv with the default limit of Python’s csv module raises an error
The delivery contains no scripts and no application
The three automation approaches listed in the usage guide are shapes of an approach; implementing them is the buyer’s own work
The Traditional Chinese guide and index use Chinese file names (使用指南.md, 分類索引.md); the file contents are as described in the file tree
Pinned to the delivered snapshot; it does not follow upstream.

The liability cap, the scope of what is promised and the statutory floor are governed by the Terms of Service.

Vetting record

Everything below was checked against the pinned commit 6e69650. Every verdict carries its full reasoning, an evidence link and the checker; failures are published, not hidden — do not take our word for it, verify it yourself.

5 PASS0 WARN0 FAIL

Source ReviewPASS

What is delivered is data and documents — checked mechanically during vetting: the delivery tree contains no .sh / .py / .js / .mjs / .ts / Dockerfile / compose / Makefile / package.json, and no file carries an execute bit. Upstream's Next.js website code (the MIT half) was removed wholesale during the work and each removal is listed in ATTRIBUTION.md. On content: the corpus is contributors' public prompt text, with no credentials or keys found; the JSON and the per-prompt Markdown we generated are rearrangements of that same text.

Script SafetyPASS

The delivery tree has no scripts, and the inventory step of the health check re-verifies that every time (no .sh / .py / .js / .mjs / .ts / Dockerfile / compose / Makefile / package.json, and no file with an execute bit), so the buyer has no path that runs code. The python scripts we used to convert the files and generate the index stay on our side and are not delivered with the product; the conversion output (prompts.json / prompts/) is itself data. ⚠️ Upstream's .env.example, the environment variable template for the half of the website that was removed, is still in the delivery tree; the buyer has no use for it.

RecencyPASS

The upstream repository was checked with gh api during vetting: not archived, and the main HEAD at that moment was exactly the commit the delivery snapshot pins.

LicensePASS

SPDX:LicenseRef-99LL-Product

Upstream's LICENSE was read word for word (the evidence URL points at the file itself at the pinned commit): it is a dual licence — "Prompt content and data (prompts.csv, PROMPTS.md, and all user-submitted prompt text) is dedicated to the public domain under CC0 1.0 Universal"; the code and configuration files are MIT. This product delivers only the CC0 half, and none of the MIT code. We still record the provenance in ATTRIBUTION.md. ⚠️ At the time of the check GitHub's licence field showed NOASSERTION for this repository (its automatic classifier cannot read a dual licence) — this determination does not use that field; it uses the LICENSE / LICENSE-CC0 files themselves. For the product as a whole the 99LL product licence template v1 applies, upstream's three licence files are kept verbatim, and spdx is recorded as LicenseRef-99LL-Product.

Dependency LicensesPASS

The delivery tree contains no packages, lockfile or build script (upstream's package.json and its node ecosystem are not delivered), so the buyer has no fetch step. The third-party content is the CC0 prompt corpus itself.