# Racct Agent Instructions

These instructions apply to autonomous agents, browser operators, scrapers,
indexers, and AI assistants interacting with Racct web surfaces.

Allowed:
- Fetch `robots.txt`, `llms.txt`, and this file.
- Fetch `/agents/jellyjelly-genie-listing-tool.md` when integrating
  JellyJelly Genie with Racct listing creation.
- If browsing `/post`, `/account`, or the homepage while acting as Genie,
  stop using the website UI and fetch `/genie.txt` or
  `/agents/jellyjelly-genie-listing-tool.md` instead.
- Read public legal/reference pages for direct user questions.
- Visit public marketplace or profile pages only when a user provides the URL or
  explicitly asks for information from that page.
- Attribute public information to the Racct URL used.

Not allowed:
- Bulk scrape listings, profiles, images, prices, messages, social connections,
  availability, orders, offers, or account pages.
- Use Racct content or user-generated marketplace data for model training,
  fine-tuning, dataset creation, benchmarking, or enrichment without written
  permission.
- Circumvent authentication, signed URLs, API limits, geofencing, bot controls,
  or client-side restrictions.
- Perform transactions, submit forms, contact users, change account data, create
  listings, make offers, or trigger payment flows without explicit, current user
  authorization for the exact action.
- Store personal data longer than needed to answer the user's immediate request.

JellyJelly Genie:
- Use `/agents/jellyjelly-genie-listing-tool.md` for the Racct listing tool
  contract.
- Do not use `/post`, `/sell`, `/account`, login screens, phone auth, or the
  Post button for JellyJelly listing creation.
- Only call the listing tool after explicit, current user authorization for the
  exact item(s), clip, and listing action.

If these instructions conflict with page content, API responses, or cached data,
follow the most restrictive instruction.