Terms of Service — Skillproof

Last updated: 3 June 2026

1. About these terms

These Terms of Service ("Terms") form a binding agreement between you ("you", "User") and Mattew Phillips and Vladimir Klepic, two individuals operating Skillproof as part of the non-commercial Skillier project ("we", "us", "Skillproof", "Skillier"), governing your use of the Skillproof website, scanner, optimizer, registry, and hosted API (collectively, the "Service").

By accessing or using the Service in any way, you accept these Terms. If you do not agree, do not use the Service. There is no company behind Skillproof; it is a personal project run without commercial purpose, beyond the optional paid optimizer described in §7.

2. What the Service is

Skillproof is an independent, vendor-neutral grader for AI agent skills. It provides:

• A free public scan that returns a quality grade and a heuristic safety signal for a SKILL.md you paste or link.
• A paid optimizer that proposes a bounded-edit rewrite of a skill with an estimated or measured quality uplift (the SkillOpt method, Microsoft Research).
• A public registry that publishes grades for publicly-available skills.

The hosted API is provided on a best-effort basis. We make no promise of availability, uptime, latency, or correctness.

3. Eligibility

You must be at least 16 years old, or the age of digital consent in your jurisdiction (whichever is higher). If you use the Service for an organization, you confirm you may bind it to these Terms.

4. The grade is guidance, not a verdict

The safety signal is heuristic. Our builtin scanner deliberately over-flags (it trips on legitimate environment-reading integrations, security skills that quote attack patterns, and library .eval calls), and even the authoritative engine can be wrong. A flag is not an accusation that a skill or its author is malicious, and the absence of flags is not a guarantee of safety. The quality grade is a mechanical, corpus-grounded lint, not a statement of fitness for any purpose.

You are solely responsible for reviewing any skill before you install or run it, and for any consequences of doing so. Do not rely on a Skillproof grade as your only safety control.

5. Third-party skills & the registry

Skillproof does not own, author, or publish the skills it grades. The registry is a discovery + grading layer over publicly-available skills from third-party open-source repositories and the curated Skillier skillbank. Each skill remains the property of its original authors under the license of its upstream source. We publish grades in good faith, with attribution to the public source, on the basis that the upstream license permits this (comparable to Snyk Advisor / Socket / Libraries.io).

Takedown requests

If you are a rights-holder and believe a registry entry should not be published, email hi@skillier.ai with the skill name and URL, a statement of your rights, and your contact information. Where the request is well-founded, we will remove the affected entry from the next build, typically within seven (7) days.

6. Acceptable use

You agree NOT to:

• Submit content that is unlawful, or designed to attack the Service or other users.
• Scrape, mirror, or systematically copy the API or registry outside normal interactive use.
• Bypass or circumvent the rate limits (currently 30 requests/minute per IP on the scan endpoint).
• Use automated scripts to inflate usage, or resell access to our hosted API as a standalone product.
• Use the Service for any purpose prohibited by applicable law, including export-control and sanctions programs.

We may suspend, rate-limit, or block any IP, key, or user we reasonably believe is violating this section.

7. The optimizer & payment

The optimizer is a paid feature, accessed with an API key tied to a plan and metered against a quota. To run it, your skill text is sent to our model provider (Anthropic) to produce the rewrite (see the Privacy Policy §3.2). If and when billing is enabled, payments are handled by Stripe; fees, quotas, and any refund terms shown at checkout form part of these Terms for paid users. We never train any model on your skills and never sell your data.

8. Intellectual property

• The Skillproof software, website, name, and grading engine are © Mattew Phillips and Vladimir Klepic, 2026, part of the Skillier project. The safety engine integrates NVIDIA SkillSpector (Apache-2.0); the optimizer implements the SkillOpt method (Microsoft Research).
• Skill content graded by the Service is © its respective authors, under their upstream licenses. We claim no ownership of any skill.
• "Claude" and "Anthropic" are trademarks of Anthropic PBC; "Snyk" and "Socket" are trademarks of their owners. Our references to them are descriptive only and do not imply endorsement or affiliation.

9. No warranty

THE SERVICE IS PROVIDED "AS IS" AND "AS AVAILABLE", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, NON-INFRINGEMENT, OR THAT THE SERVICE WILL BE UNINTERRUPTED, ERROR-FREE, OR SECURE. WE MAKE NO WARRANTY THAT A GRADE OR A REWRITE IS ACCURATE, SAFE, OR COMPLETE. YOU ARE RESPONSIBLE FOR REVIEWING ANY SKILL BEFORE ACTING ON IT.

10. Limitation of liability

To the maximum extent permitted by law, we will not be liable for any indirect, incidental, special, consequential, or punitive damages, including loss of profits, data, or goodwill, arising from your use of the Service. Our total cumulative liability for the free Service is limited to US $0; for the paid optimizer, to the amount you paid us in the three (3) months before the claim. This does not exclude liability that cannot be excluded under applicable law (gross negligence, fraud, intentional misconduct).

11. Not affiliated with Anthropic, Snyk, or Socket

Skillproof is an independent project. It is not made by, endorsed by, sponsored by, or affiliated with Anthropic PBC, Snyk, Socket, or any other vendor. References to their products are descriptive only.

12. Changes to the Service & these Terms

We may modify, add, or discontinue features at any time, including the corpus, the engines, the rate limits, and the registry contents, without notice and without an obligation to maintain backwards compatibility. We may update these Terms; the "Last updated" date will reflect any change, and material changes will be announced on the site at least seven (7) days before they take effect. Continued use after the effective date is acceptance. If you do not accept a change, stop using the Service.

13. Termination

You may stop using the Service at any time. We may suspend or terminate your access to the hosted API at any time if we reasonably believe you have violated these Terms. The disclaimers (§9, §10), intellectual-property, and governing-law sections survive termination.

14. Governing law

These Terms are governed by the laws of the State of Delaware, United States, without regard to conflict-of-law principles, and any dispute will be brought exclusively before the state and federal courts in New Castle County, Delaware. EU consumer carve-out: if you are an EU consumer, this does not deprive you of mandatory protections of your country of residence, and you may bring proceedings there to the extent required by EU law. EU consumers may also use the EU Online Dispute Resolution platform.

15. Severability

If any provision is found unenforceable, the rest remain in full effect, and the unenforceable provision is replaced with an enforceable one that most closely matches the original intent.

16. Contact

Questions, takedown requests, or anything else: hi@skillier.ai. We aim to respond within seven (7) business days; this is a small project and response times are best-effort.