Editorial Independence: What Payment Cannot Change in an Agent Skill Verdict

Agent Skill verdicts on Verify AI Skills are not sponsored. Payment cannot change a score, label, safety grade, confidence level or hub ranking, and sponsored placements are labelled. Editorial independence means that commercial work stays separate from sourced facts and verdicts. The site accepts disclosed advertising, newsletter sponsorship and custom skill-building work.

Can payment change an Agent Skill verdict?

No. Payment cannot change an Agent Skill verdict's score, label, safety grade or confidence level. It cannot hide a known failure, remove criticism or lift a skill's position on a best-skills hub.

A verdict comes from evidence, and money is not evidence. No sponsor, advertiser or client can buy any of these 4 outcomes:

• Change a score, verdict label, safety grade or confidence level.
• Hide a known failure.
• Remove legitimate criticism from a verdict page.
• Improve a skill's position on a best-skills hub.

The same rule covers every published verdict. A paying partner reads the same page as everyone else.

What can commercial partners buy?

Commercial partners can buy clearly labelled sponsorship placements on the site and in the newsletter. They can buy custom skill-building work with an agreed scope. Sponsored material always looks different from editorial content.

Commercial partners can buy 2 things:

• Sponsorship placements, on the site or in the newsletter, each one clearly labelled.
• Custom skill-building work, done to a scope both sides agree in advance.

Sponsored material is visually distinct from editorial content, so a reader can tell a paid placement from a verdict.

How is sponsored and affiliate content labelled?

Sponsored content is labelled where it appears and is visually distinct from editorial content. Affiliate relationships, once introduced, are disclosed next to the content they affect. The site has none at launch.

A sponsored message shows its label in the same place the reader sees it. Readers never need to open a separate page to learn that a placement was paid for. An affiliate link, once one exists, gets its disclosure beside the link itself. Today the site has no affiliate relationships, so no verdict page carries one.

How are best-skills hub rankings kept independent?

Best-skills hub rankings are kept independent by following published verdict scores only. Ties are broken by confidence, safety grade and name. No payment or sponsorship moves a skill up or down a hub.

Each of the best-skills hubs ranks only skills whose main job matches the hub's task. Order comes from the published score. Two skills with the same score are split by confidence level first, then safety grade, then name. Sponsorship has no input at any step. A sponsor's own skill sits wherever its score puts it.

The full rule set sits on the page explaining how verdicts are scored and ranked. Anyone can check a hub's order against the scores on each verdict page.

What can skill publishers do about their verdict?

Skill publishers can submit documentation, fixes and new evidence about their verdict through a correction request. They never get final editorial approval. Their evidence is weighed under the same published rules as every other source.

A skill publisher is the person or company that maintains a skill's repository. Publishers have 3 routes:

• Submit documentation that explains how the skill works.
• Report fixes that close a failure named in the verdict.
• Send new public evidence, dated and linked.

Anyone can submit a correction with a source, and publishers use the same route. Publisher statements and benchmarks stay labelled as publisher claims, never as independent evidence.

How is a skill from the custom-skills service disclosed?

A skill built by our custom-skills service is disclosed on its own verdict page. Payment for that build cannot change its score, label, safety grade, confidence level or hub position.

The custom-skills service is paid work where we build an Agent Skill for a client. A verdict on such a skill states that link on its page. Every limit in the table above applies to it.

How is AI used to produce Agent Skill verdicts?

We use AI to collect evidence and draft each Agent Skill verdict from a fixed checklist. Automated checks catch rule breaks, and a human editor reviews every verdict before it is published.

Each draft is based on the skill's files at a pinned commit and dated public evidence. Automated checks test every draft, and a failed check blocks it. A human editor reviews every verdict before it is published.

The checklist and the automated checks add 4 rules:

• Search for negative evidence on every skill, not only praise.
• Cite only sources that were opened and read, dated inside the window from 2026-03-30 to 2026-09-30.
• Reject promotional phrases such as "the best", "leading" and "most popular".
• Match each verdict label to its score band, with every cited URL returning HTTP 200.

Our methodology lists each step.

How were the reviewed Agent Skills chosen?

The reviewed Agent Skills were chosen from the agenticskills.io catalogue of listed skills. Most are popular listed skills, so the set does not cover the long tail. Each one follows the same published rules.

The agenticskills.io catalogue is a public directory of Agent Skills, and every skill here came from its listings. Every skill was reviewed on 2026-09-30.

Common questions

Are Agent Skill verdicts sponsored?

No. Verdicts carry no sponsorship, and any sponsored placement on the site is labelled. Browse every published verdict to see the format.

Can a publisher pay to remove a negative finding?

No. Payment cannot hide a known failure. Factual errors go through the corrections process with a source.

Does a custom-built skill get a better verdict?

No. A skill from the custom-skills service is disclosed on its page and scored under the same rules.

Are agent skills useful at all?

Yes, many are. Of the skills we have reviewed, 10 are Recommended and 10 are Recommended with caveats.