BlackFriday-GPTs-Prompts: a category-sorted list of free GPT prompts, with an Android agent APK bolted on top

What the repository actually contains

This is a curated index, not a program. The top level holds one Markdown file per subject: Programming.md, Marketing.md, Academic.md, Job-Hunting.md, Game.md, Creative.md, Prompt-Engineering.md, Business.md, Productivity-&-life-style.md, Jailbreaks.md, Models.md and 2024-May-Update.md. There is also a gpts/ directory and a LICENSE file. The README is essentially a table of contents pointing at those files, with the repository description framing the whole thing as a list of free GPTs that do not require a Plus subscription.

The audience follows from that shape. Someone who has a ChatGPT account without a paid plan, wants a system prompt for a specific task, and is willing to copy it by hand. The value is the sorting: instead of searching a forum thread, you open Programming.md or Academic.md and read. The cost is that a Markdown list is only as good as its links, and the README gives no maintenance promise for the individual entries.

One thing to notice before anything else. The README does not begin with the prompt list. It begins with a centered block advertising AnyClaw, an OpenClaw plus Codex Android app hosted at friuns2.github.io/openclaw-android-assistant, with two screenshots and a Download APK link. The prompt index starts below that. The homepage field on the repository points at the same app page, not at the prompt collection. If you arrived expecting prompts, that is the first mismatch to resolve.

How the index is organized and how you read it

There is no build step, no schema and no manifest. The mechanism is file-per-category plus a README that links to each file. In the README the links are relative paths such as ./Programming.md and ./Jailbreaks.md, so they resolve on GitHub and in a local clone alike.

That means the data flow is entirely manual: a reader opens the README, picks a category, opens the file, copies a prompt, and pastes it into whatever chat interface they use. Nothing is fetched at runtime. There is no index file listing every prompt with an ID, and the repository does not state a format for entries inside the category files. If you want to consume this programmatically you would be parsing Markdown by hand, and the repository does not document any stable structure to parse against.

A second directory, gpts/, sits alongside the Markdown files. The README does not describe what it holds, and the repository does not document a naming convention for it. Treat that as unknown until you open it.

The practical consequence: this is a reading list. Its quality is a function of the links and prompts inside each file, which the README does not summarize or date. The only dated artifact visible is 2024-May-Update.md, which suggests the collection had a snapshot in May 2024. That is the closest thing to a freshness signal in the repository.

Reading the prompt files without installing anything

There is no package, no installer and no CLI. The README's only action links are Download APK and GitHub, and the APK belongs to the separate Android app, not to the prompt collection. So the way to obtain the prompts is to clone or download the repository itself.

The README gives no git command, so the only concrete steps available are the ones the repository itself provides: the GitHub link and the list of Markdown files. Open the repository on GitHub and click the file for the category you want, or use the Clone or download button that GitHub shows on the repository page. The README's own example of a category link is this relative path:

./Programming.md

On GitHub that resolves to the Programming category file, and the same relative path works from a local copy of the repository. The README lists the other categories in the same form, including ./Marketing.md, ./Academic.md, ./Job-Hunting.md, ./Game.md, ./Creative.md, ./Prompt-Engineering.md, ./Business.md, ./Productivity-&-life-style.md and ./Jailbreaks.md.

A first real use: open ./Programming.md, copy a prompt you want, and paste it as the instructions field when you create a custom GPT, or as the first message in a chat. The repository does not document recommended models per prompt, so model choice is yours.

The Jailbreaks file and the limits of a copied prompt

Jailbreaks.md is listed as a top-level category like any other, and the repository topics include jailbreak and exploit alongside prompts and spam. That is a deliberate part of the collection, not an accident, and it is the part most likely to break without warning. Prompts written to get around a model's refusals are the ones most sensitive to model updates, and the README offers no compatibility notes, no model version per entry and no report of which ones still behave as described.

Even for the ordinary categories, a copied prompt is a snapshot. The repository does not version individual prompts, does not record when an entry was added, and provides no test or example output. If a prompt stops producing what you expect, there is no changelog to check and no issue template described in the README to report it against. You can compare your copy against the current file on GitHub to see what changed, but you cannot tell why.

This is the case where it is the wrong tool. If you need prompts that are evaluated, versioned and tied to a specific model revision, a Markdown list maintained by hand is the wrong container. If you need a single prompt to do one job and you are willing to iterate on it yourself, the list is a reasonable starting point and the cost of a bad entry is one wasted paste.

Where it sits next to prompt hubs like Awesome ChatGPT Prompts

The obvious comparison is f/awesome-chatgpt-prompts, which is the other widely referenced free prompt collection on GitHub. The difference is in the container, not the idea. That project is built around a single prompts.csv as the machine-readable source, with the README generated from it, which makes it possible to consume the whole set as data and to submit one row per prompt. This repository is organized as one Markdown file per human category, with the README acting as a hand-written table of contents.

The trade-off is concrete. Category files are easier to browse when you do not know what you are looking for: you open Marketing.md and read. A CSV is easier to filter, diff and feed into a script. If your workflow is "find a prompt for a task I can name", the category split here is pleasant. If your workflow is "load every prompt into a tool", the absence of a structured source means writing a parser against unstructured Markdown, and the repository does not promise the format is stable.

A second difference in scope: this repository also carries Jailbreaks.md and a gpts/ directory. The former is a category the CSV-style hubs generally do not maintain as a first-class file. That is a real distinction in what you get, and also in what you take on.

Licence, maintenance and what upgrading costs you

The repository is MIT licensed, and LICENSE is present at the top level. MIT is permissive: it allows use, modification and redistribution provided the copyright notice and permission notice are retained. Note what that does and does not cover. The licence text governs the repository contents. It does not grant you rights to the third-party GPTs, sites or services that the category files link to, and it does not speak to the terms of the model you paste a prompt into. Those are separate questions, and the repository does not address them. This is a description of the licence file, not legal advice.

Maintenance is a separate matter from the licence. The last push to the repository was on 2026-03-18, so the repository is not archived but is not showing recent activity either. The only dated artifact inside the collection is 2024-May-Update.md. Anyone depending on a specific entry should check that file's history rather than assume the whole collection is current.

Upgrade cost is low by construction. There are no dependencies to bump, no build to run and no API surface to track. The cost is review time: when you pull, you are reading a Markdown diff and deciding which changed prompts you care about, with no version tags and no release notes to guide you. The repository shows no releases, so there is no release channel to subscribe to.

The AnyClaw APK in the README is a different project

The most confusing part of this repository for a new reader is the top of its README. It advertises AnyClaw, described there as OpenClaw plus Codex running natively on Android, in one APK, with no root and no Termux, linking to a separate repository and to a download page under friuns2.github.io. The repository's homepage field points at that same Android page.

Nothing connects the APK to the prompt files beyond the same owner and the same README. The prompt index is the body of the repository; AnyClaw is a separate codebase with its own repository and its own download. If you came for prompts, ignore that block. If you came for the Android agent, the APK page is where the README sends you, and this repository is not the place to read about how it works, because the README does not describe it beyond the tagline and the screenshots.

That arrangement is worth flagging as a documentation problem rather than a technical one. A visitor landing on the GitHub page sees an app promotion above the actual content, and the homepage link reinforces the impression that this repository is about the app. The prompt collection is real and it is what the files contain, but you have to scroll past the advertisement to reach it.

Editorial conclusion

Adopt it if you want a plain Markdown prompt library you can clone and read offline, or if you want a starting set of system prompts to paste into a custom GPT. Skip it if you need versioned prompts, evaluation data, or anything with an API. Before relying on it, open the file for your category and check whether it links out to third-party GPTs, because those links age independently of the repository, and read LICENSE to confirm what the MIT grant covers for you.

Frequently asked questions

Official sources

1. friuns2/BlackFriday-GPTs-Prompts on GitHub
2. Issues
3. License: MIT
4. Project website
5. README