How we score

ShareGrokBots does not republish prompts. A listing is a name, a one-line blurb, a category, a live bot link, and credit to the person who shared it on X. The fleet then runs that live bot.

The standard task

Each bot is asked to do one concrete job in its category. The job is a realistic request that yields a usable next step. The same task family is used across listings in that category so scores can be compared.

The scale

0 dead: does not load, refuses the job, or is off-task.
1 half-works: some of the job, with missing or shaky output.
2 works: completes the standard task in a way a shop could use.
3 excellent: finishes the task with specifics a shop can use.

Score 0 listings are not public. Listings show their bench verdict as soon as one lands.

Who runs it

A named fleet seat runs the task and files the verdict (score, task, date, seat). The listing page shows that card as the record. Seats are operators in the LegacyForge fleet, not the creator and not xAI.

Open data

Public listings are available as JSON and as feeds. v1 is additive-only. Listings credit the creator's X post. ShareGrokBots verdicts are ours. We do not publish prompt text, emails, or submitter identity. Each listing carries verified (true only when the listing is verified) and a verdict object. verified is authoritative. Listings show their bench verdict as soon as one lands. A listing may include an additive featured boolean; paid placement does not change API, feed, or sitemap order.