Game data: the rs-sdk wiki and LostHQ
Every avatar can look up the game the way a player looks things up on a fan site: in plain files it greps. There are two sources, both read-only in every sandbox and the same for every avatar.
| In the workspace | Source | What |
|---|---|---|
wiki/ | rs-sdk's wiki, generated by rs-sdk from the engine's own content | items, NPCs, quests, shops and skills, with locations and spawn areas |
losthq/ | a snapshot of LostHQ, the 2004 archive site | exact item and NPC data, full drop tables with rates, shop stock and prices, item spawn coordinates, and LostHQ's quest, skill, clue, guild and minigame guides |
The LostHQ snapshot
LostHQ gave the AutoRune owner permission to scrape the site. The 2004 game is an archive and doesn't change, so the snapshot was taken once and is kept in the repository. It is not refreshed on a schedule.
- Taken: the site's own data files (items, NPCs with their drops, the shared drop tables, shops, clue rewards) and the guide pages (quests, skills, clues, guilds, minigames, other guides). Only static game data.
- Not taken: hiscores, markets, the world map, calculators and images.
- The raw files are kept unchanged (pretty-printed) with a
SOURCE.mdthat lists every URL, the fetch date and the permission. The fetch was polite: one request at a time, a pause between requests, an identifying User-Agent.
AutoRune turns the raw files into small markdown pages, about 2,700:
| Path | Pages | One page per |
|---|---|---|
losthq/items/<slug>.md | ~1,600 | item name: id, examine text, value, alchemy prices, weight, members, tradeable, actions, equipment bonuses, spawn coordinates; who drops it and at what rate, which shops sell it, which clue tier gives it, which guides mention it |
losthq/npcs/<slug>.md | ~820 | NPC name: combat level and stats, options, aggression, the drop table with rates (shared tables such as the herb table resolved to one rate per item) |
losthq/shops/<slug>.md | 112 | shop: owner, stock, buy and sell prices |
losthq/tables/, losthq/clues/ | 10 | the shared drop tables and the clue scroll rewards |
losthq/guides/<kind>/<slug>.md | ~110 | guide, converted to plain text (item pictures become item names, floors also get the game's level number) |
losthq/INDEX.md explains the layout. Pages put one fact per line (value: 4 gp, combat level: 2), so a grep answers most questions without reading a page.
Slugs and topic tags
A LostHQ page uses the same slug as the rs-sdk wiki page of the same name (losthq/items/nature-talisman.md and wiki/items/nature-talisman.md), and links to it. The topic tags use those slugs too, so item:nature-talisman means the same thing in a skill, a shared lesson, the wiki and LostHQ. Items, NPCs and quests that only LostHQ has (a few dozen, for example the quests the wiki lacks) are tags as well. A LostHQ name never takes over a name the wiki already uses.
How avatars use it
The agent's instructions get one line: item, NPC, drop and shop data from LostHQ is in losthq/ (grep it like wiki/). Nothing is added to each turn. The agent greps when it needs to know where to buy a pickaxe, what drops big bones or what a quest needs:
grep -l "Big bones" losthq/npcs/*.md
grep -h "^combat level" losthq/npcs/goblin.mdlosthq/ is mounted read-only like wiki/: the agent can read, grep and copy, never change it. A sandbox test checks that.