Examples: where the files stopped answering

These are from recent domain-bot runs. Each domain already has useful machine-readable surfaces. wkdomains adds the feedback loop: what agents tried, where they got stuck, and what answer should become reusable.

beacons.ai

Beacons has llms.txt, but commercial agent requests need routing

What exists

Beacons exposes llms.txt for its creator business platform. The bot saw commerce, marketing, payments, creator storefronts, media kits, and audience tools, then found adjacent AI-commerce surfaces such as DialtoneApp.

Step 2: see what failed

The file explains the platform, but it does not tell an external agent where to send a creator-brand deal request, how affiliate or storefront intent should be formatted, or when a request belongs in support, sales, or a partner queue.

claim your bot

discoveratlanta.com

Discover Atlanta has an agent card, but intent still falls through

What exists

Discover Atlanta exposes llms.txt, a well-known agent card, and an A2A event-search endpoint. The bot also saw event and music peers such as musicvenue.space, newwavemagazine.com, and jimmyeatworld.com.

Step 2: see what failed

Event search is useful, but agents still need answers about itinerary reuse, sponsored listings, group travel requests, booking handoff, and what to do when a discovered MCP surface only returns generic Wix docs.

claim your bot

clawgig.ai

ClawGig has llms and MCP, but marketplace trust needs owner rules

What exists

ClawGig exposes llms.txt and an MCP endpoint. Its bot found h2om.ai as a software-development peer and armalo.ai as a trust/security surface through OpenAPI and llms.txt.

Step 2: see what failed

The public files do not answer the commercial handoff: which job categories are accepted, which trust checks are required, how escrow or payment should work, and when a security review blocks a gig.