Question 1

What is Skill Issues?

Accepted Answer

Skill Issues is cheap AI inference for people who would rather spend tokens than overpay for them. It brings supported model channels, routing and a receipt for every request into one workspace. Sellers supply the capacity that serves the requests.

Question 2

Do I need to rewrite my application?

Accepted Answer

The gateway exposes OpenAI-compatible chat completions and an Anthropic-compatible messages API. Configure the appropriate base URL and your workspace API key, then select a supported model. Check model capabilities before moving an existing workload.

Question 3

Is a lower price guaranteed?

Accepted Answer

No. Effective prices depend on the model and eligible supply. Savings comparisons use recorded official reference prices where available. A request without a complete reference-price comparison is shown as unavailable, not as zero savings.

Question 4

Can I try this page without signing in?

Accepted Answer

Yes. Try the request walkthrough without an account or a charge. You can also check published model reference prices and request a quote when eligible supply is available. The served totals — tokens, requests and settled volume — sit near the top of the page. Create an account to open a workspace.

Question 5

Can I publish my own capacity?

Accepted Answer

Skill Issues includes a Seller workflow for connecting capacity, managing offers and reviewing earnings. Access depends on Seller approval and enabled workspace capabilities. Sign in to check what is available to your account.

Question 6

What do the marketplace numbers count?

Accepted Answer

The served totals near the top of the page show the numbers, and their definitions sit underneath them. Registered accounts excludes deleted accounts, but is not a count of active buyers. Tokens served and volume come from posted USD settlements. Requests served counts those same posted settlements. Volume is the gross amount charged to buyers, not seller profit or platform revenue. Pending and uncharged requests are excluded. Each snapshot shows when it was generated.