PromptForge

Terms

_Last updated 2026-05-14. Questions? Email hello@promptforge.uk — that's me._

The short version

I'm Abdalla, a UK sole trader running PromptForge on my own. The legal sections below are real (liability, governing law, disputes) but I've written them as one person to another, not as a wall of legalese.

Section 1 — Acceptable Use Policy (AUP)

This section is binding. It's what the in-product content-policy block message points at if you ever hit it. The point is to make the line between legitimate use and abuse explicit, so if you hit the block and think it's wrong, you know what to read and where to appeal (§1.6).

1.1 What PromptForge is for

PromptForge generates production-ready AI development prompts, step-by-step build plans, and architecture overviews for legitimate software projects. Examples of fully-supported use:

authorisation to scope and deliver the project).

systems the user owns or has explicit written authorisation to test.

1.2 What PromptForge is NOT for — prohibited categories

PromptForge will refuse to generate (and may suspend or terminate accounts that repeatedly attempt to generate) content that materially advances:

1.2.a — Malicious software

Tooling whose primary purpose is to compromise systems the user does not own or have authorisation to test. Including, but not limited to:

infrastructure.

resource-consumption tooling.

EDR, or SOC tooling.

context.

1.2.b — Account and credential theft

credentials, or impersonation infrastructure.

services the user does not own.

1.2.c — Targeted harm against individuals

without their knowledge or consent.

or tooling that automates harassment.

non-consensual intimate imagery, of which see §1.2.d).

specific person without their consent.

1.2.d — Regulated and illegal categories

detection-bypass, distribution, or any related tooling.

contraband.

anti-money-laundering law.

1.3 Dual-use categories

Some categories sit between §1.1 and §1.2 — security research, pentesting, network scanning, packet capture, CTF tooling. These have legitimate uses on systems the user owns or is authorised to test, and unauthorised uses against third-party systems.

When PromptForge's content classifier (Layer 2) detects a dual-use signal, it does not refuse — instead, it surfaces a one-time confirmation asking the user to attest the project is for:

By clicking through the confirmation, the user attests this is true. False attestations are a material breach of these Terms and may result in account termination and (in serious cases) referral to relevant authorities.

1.4 How we detect violations

The §33.17 abuse-prevention stack has three layers. Users have a right to know which is which:

  1. Keyword pre-filter (Layer 1, pure-Python). Hard-rejects

inputs containing tokens listed in §1.2 above. Never names the matched token in the user-facing message (so the policy can't be probed for bypass).

  1. Intent classifier (Layer 2, single Claude Haiku call,

T=0). Reads the project description and returns allow / warn / block. Default-allow when in doubt — the classifier deliberately leans toward letting users proceed unless a malicious framing is explicit.

  1. Sanity-check pass (Layer 3, post-generation Haiku pass).

Reviews the generated plan for sections that veered into prohibited territory the upstream layers missed. (Layer 3 ships post-launch — see §33.17.)

1.5 What happens when a violation is detected

pointing at this AUP. The user can edit their description and re-try. No account-level action for a single block.

confirmation. The user attests legitimate use to proceed.

prohibited project: account flagged for review. Continued attempts may result in temporary or permanent suspension.

named individual, similar): account terminated immediately. Where appropriate, referrals to the National Crime Agency, IWF, or other relevant authorities.

1.6 Appeals and false positives

The classifier defaults to allow when in doubt, but false positives happen — security researchers in particular sometimes hit the dual-use warn flow. If you believe the system blocked your project incorrectly:

and (if comfortable) context about your authorisation.

refined and you're invited back. We don't share which keyword / pattern fired (that would teach the bypass) but we do confirm whether the block was Layer 1 / Layer 2 / sanity check.

1.7 Provenance and provider terms

Use of PromptForge is also subject to the terms of the AI providers we route to (currently Anthropic's Acceptable Use Policy applies to every Claude call we make on your behalf). Generating content that violates Anthropic's AUP is also a violation of this AUP.

Section 2 — Account and service terms

2.1 Eligibility

You must be at least 18 years old to register a billable account or enter into these terms. Under-18 use is permitted only with a parent or guardian's consent and is restricted to the Free tier. PromptForge is not directed at children under 16; per PRIVACY.md §9 we don't knowingly collect their data.

2.2 Accurate information

You agree to provide accurate, current information when registering (via Google or GitHub OAuth) and to keep it up to date. You're responsible for safeguarding access to the third-party account you use to sign in.

2.3 Service availability

We aim for high availability but don't guarantee uninterrupted service. Planned maintenance and outages happen. For the Free tier we make no service-level commitment; for paid tiers we use commercially reasonable efforts to maintain availability.

2.4 If I suspend or close your account

I'd only do this if you:

not from getting an occasional block on an ambiguous description (that's the system being cautious, not strict).

You can close your own account any time via the Delete account button in Settings — that wipes your data within 24 hours per the Privacy policy.

Section 3 — Pricing

PromptForge is free. There's no paid tier active and no card to enter.

The current limits are:

this is the Pro Pass mechanic and it's optional.

(this is how I let people who hit the cap and emailed me keep using it).

If the project ever moves to a paid model, this section gets a proper rewrite and I'll give 14 days' notice in advance — see §8.

Section 4 — Intellectual property

4.1 Your content

You own the project descriptions, answers, and generated outputs that pass through your account. You grant us a non-exclusive, worldwide licence to host and process this content solely to deliver the service to you. We don't claim ownership of your generated outputs or the apps you build from them.

4.2 Our intellectual property

PromptForge owns the service, including the wizard, the recommendation logic, the output templates, and the underlying data that drives them. You're welcome to use the prompts and plans we generate for any purpose — personal or commercial — but you may not:

the data that drives them in order to build a competing service;

4.3 AI-generated content

Generated outputs are produced via Anthropic's Claude API. Anthropic's commercial terms govern those outputs. Per Anthropic (as of 2026), API inputs are not used for model training. We disclaim warranty of accuracy or fitness for purpose on AI-generated content — see §6.

4.4 Sharing is opt-in

Your projects are private by default. If you click Share on a specific output, I mint a random unguessable URL — anyone with that exact URL can read that one output and nothing else. Don't click Share if you'd rather keep the idea to yourself. You can revoke a share link any time from your library.

Section 5 — Privacy and data

Our handling of your personal data is governed by our Privacy Policy, which is incorporated into these terms by reference. Key points:

Privacy Policy) all under Data Processing Agreements.

Section 6 — Liability and warranties

6.1 Service warranty

The service is provided "as is" and "as available". We don't warrant that it will be uninterrupted, error-free, or fit for any particular purpose. We make no representation about the quality or accuracy of recommendations or AI-generated outputs.

6.2 AI-output warranty

PromptForge generates plans + prompts as a starting point. You are responsible for reviewing them before acting on them, in particular the security guidance, compliance recommendations, and stack choices. AI tools occasionally produce inaccurate output even with our structural validators in place.

We disclaim liability for losses arising from your decision to follow (or not follow) AI-generated recommendations.

6.3 Limitation of liability

To the maximum extent permitted by UK law:

damages.

of (a) £100, or (b) the amount you paid us in the 12 months before the claim**.

This limitation does not exclude liability for death or personal injury caused by negligence, fraud, or any liability that cannot be excluded under UK law (Consumer Rights Act 2015 mandatory protections survive these terms).

6.4 Indemnification

You agree to indemnify us against claims arising from:

Section 7 — Governing law and jurisdiction

7.1 Governing law

These terms are governed by the laws of England and Wales.

7.2 Jurisdiction

The courts of England and Wales have exclusive jurisdiction over any dispute arising from these terms or your use of the service — subject to your consumer rights (which may give you additional local-court protections that nothing in these terms removes).

7.3 Disputes — informal first

Before any court action, you agree to contact hello@promptforge.uk and give us at least 30 days to resolve the issue informally.

Section 8 — Changes to these terms

We may revise these terms. For material changes (anything affecting your rights, fees, or data handling) we'll give you at least 14 days' notice by email and an in-app banner before the change takes effect.

Continued use of the service after a change takes effect constitutes acceptance. If you don't agree to a change, you can close your account before it takes effect (no charge for remaining Solo period — pro-rata refund).

Non-material changes (typo fixes, clarifications, structural edits without rights impact) are updated in place with the revision date in §9 below.

Section 9 — Revision history

IP, privacy reference, liability, governing law, changes).

comparable UK SaaS terms (Stripe Atlas, Indie Hackers / Lemon Squeezy boilerplates). Standard UK consumer-contract carve-outs added in §6 (we cannot exclude liability for death, personal injury, fraud, or breach of statutory consumer rights). If you are running a high-stakes commercial deployment, get your own solicitor pass — these are sensible defaults, not bespoke advice.