Token导航 LogoToken导航TokenDH.com
研究检索只读github未标认证来源可访问许可证需确认审计提醒

roomroom 搜索

Agent Skill

room 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

321

周安装

13

GitHub Stars

37

下载量

101
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:room(room 搜索)
来源仓库:https://github.com/simhacker/moollm
仓库路径:skills/room
安装命令:
npx skills add https://github.com/simhacker/moollm --skill room
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/simhacker/moollm --skill room

简介

room 用于查找、检索和筛选相关信息。

  • 适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词快速定位结果。
  • 通过 npx skills add 命令从指定仓库安装并使用。
  • 建议确认权限范围和维护状态,注意是否触发联网或文件操作。
  • room 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Room

Rooms are intertwingled navigable activation context maps. Entering = calling. Exiting = returning.

Directories as cognitive spaces where cards come to life.

The Metaphor

FilesystemSimulationProgramming
DirectoryRoomStack frame
cd room/EnterFunction call
cd..ExitReturn
Files in dirActive entitiesLocal variables
LinksExits/doorsCalls to other functions

Room Anatomy

room-name/
├── ROOM.yml        # Room definition (REQUIRED!)
├── README.md       # Room's voice (optional)
├── CARD.yml        # Card sidecar (optional)
├── character-*.yml # NPCs embedded in room
├── object-*.yml    # Objects in room
├── inbox/          # Objects thrown INTO this room
├── outbox/         # Objects staged to throw OUT
├── region/         # Sub-region of this room
│   └── ROOM.yml    # Region must also declare itself!
└── nested-room/    # Full sub-room (different location)
    └── ROOM.yml

Regions vs Sub-Rooms

What's a Region?

A region is a sub-directory that represents a portion of the same room — like the stage area of the pub, or the back corner of a library.

# pub/ROOM.yml
room:
  name: "The Gezelligheid Grotto Pub"

  # Regions are PARTS of this room
  regions:
    stage:
      path: "stage/"
      description: "The performance stage"
      visibility: public

    back-room:
      path: "back-room/"
      description: "Private back room"
      visibility: private
      requires: "bartender approval"

    bar:
      path: "bar/"
      description: "The bar counter area"

Regions Have Rules

Each region can have its own:

# pub/back-room/ROOM.yml
room:
  name: "Back Room"
  type: region                    # Marks as region, not full room
  parent: "../"                   # Part of parent room

  # Access control
  access:
    visibility: private           # Not visible to everyone
    requires: "bartender approval OR staff badge"
    who_allowed:
      - "characters/staff/*"
      - "player if has_flag('vip_access')"
    who_denied:
      - "characters/troublemakers/*"

  # Ethics & behavior
  rules:
    - "No recording"
    - "Confidential conversations"
    - "Staff only by default"

  # Privacy
  privacy:
    eavesdropping: false          # Can't hear from outside
    visible_from_parent: false    # Can't see inside from pub

  # What happens in the back room...
  narrative:
    on_enter: "The door closes behind you with a soft click."
    on_exit: "You return to the bustling pub."

Visibility Types

TypeDescription
publicAnyone can see and enter
visibleCan see but may need permission to enter
privateHidden unless you know about it
secretHidden AND requires discovery

Region vs Full Sub-Room

FeatureRegionSub-Room
Part of parent?YesNo
Own identity?PartialFull
Exit returns to?ParentVaries
Shares parent context?YesNo
Type fieldtype: regiontype: room

Directory Type Declaration

The Rule

Every directory in an adventure MUST declare what it is.

Directory TypeDeclaration File
RoomROOM.yml
RegionROOM.yml (with type: region)
CharacterCHARACTER.yml
Adventure rootADVENTURE.yml
PersonasROOM.yml (with type: personas)
StorageROOM.yml (with type: storage)

Lint Error: Undeclared Directory

# LINT ERROR: Directory without type declaration
- type: MISSING_TYPE_DECLARATION
  severity: WARNING
  path: "pub/mysterious-corner/"
  message: "Directory has no ROOM.yml, CHARACTER.yml, or other type declaration"
  suggestion: "Add ROOM.yml with appropriate type field"

Valid Non-Room Directories

Some directories aren't rooms and that's OK:

# These don't need ROOM.yml:
messages/           # Mail storage (system)
inbox/              # Postal inbox (system)
outbox/             # Postal outbox (system)
sessions/           # Session logs (meta)
images/             # Asset storage (meta)

Marking System Directories

# Alternative: mark as system directory
# pub/messages/.meta.yml
meta:
  type: system
  purpose: "Mail storage for pub"
  requires_room_yml: false

Container Directories (Inheritance Scopes)

Some directories are inheritance containers — they provide shared properties to child rooms without being navigable themselves. Like OpenLaszlo's <node> element.

# maze/CONTAINER.yml — Not a room, but defines inherited properties
container:
  name: "The Twisty Maze"
  description: "Groups maze rooms with shared grue rules"

  inherits:
    is_dark: true
    is_dangerous: true
    grue_rules:
      can_appear: true

  ambient:
    sound: "dripping water"
    light_level: 0

All child rooms (room-a/, room-b/, etc.) automatically inherit these properties!

Alternatively, you can make the container into an actual room:

# maze/ROOM.yml — The maze entrance IS a room
room:
  name: "Maze Entrance"
  description: "Dark passages branch off in every direction..."

  exits:
    a: room-a/
    b: room-b/
    # ... etc

DESIGN CHOICE:

  • Use CONTAINER.yml if you want inheritance without navigation (see container skill)
  • Use ROOM.yml if you want the directory to be a navigable space

Hierarchy Example

adventure-4/
├── ADVENTURE.yml           # Adventure declaration
├── pub/
│   ├── ROOM.yml            # Room declaration
│   ├── stage/
│   │   └── ROOM.yml        # Region (type: region)
│   ├── bar/
│   │   └── ROOM.yml        # Region
│   ├── back-room/
│   │   └── ROOM.yml        # Private region
│   └── messages/
│       └── .meta.yml       # System directory (no ROOM.yml needed)
├── characters/
│   ├── ROOM.yml            # Hall of characters (type: personas)
│   └── don-hopkins/
│       └── CHARACTER.yml   # Character declaration
└── maze/
    ├── ROOM.yml            # Room declaration
    └── room-a/
        └── ROOM.yml        # Sub-room (full room, not region)

ROOM.yml Structure

room:
  name: "Debug Session"
  purpose: "Hunt down the authentication bug"

  context:
    - "Bug: Login fails with valid credentials"
    - "Suspected: Session cookie handling"

  cards_in_play:
    - instance: "goblin-001"
      card: "Git Goblin"
      goal: "Find when bug was introduced"

  working_set:
    - "ROOM.yml"
    - "state/progress.yml"

  exits:
    parent: "../"
    related: "../feature-work/"

  # Optional: position in 2D world-space
  world_position:
    x: 5
    y: 12

  # Optional: objects with positions in room-space
  objects:
    - name: "workbench"
      position: {x: 3, y: 7}

Spatial Coordinates

Rooms can exist in world-space. Objects can have positions in room-space.

# World-space: where is this room in the world?
world_position:
  x: 5
  y: 12

# Room-space: where are objects within this room?
objects:
  - name: "workbench"
    position: {x: 3, y: 7}

Navigation can use coordinates:

  • NORTH from (5,12) → find room at (5,13)
  • Named exits override coordinates

Not all rooms need coordinates. Abstract spaces can exist outside world-space.

Vehicles: Portable Rooms That Move

A vehicle is a room you can embark, drive, and disembark.

# vehicle-tent.yml
room:
  name: "Research Tent"
  is_vehicle: true
  world_position: {x: 5, y: 12}  # Changes when you drive
CommandEffect
EMBARK tentEnter the vehicle room
DISEMBARKExit to current world location
DRIVE NORTHMove vehicle (and occupants) to (5,13)

Riding the Turtle

RIDE the turtle. Move around the room, draw on the floor, jump through doors:

> RIDE turtle
You mount the turtle. The world scrolls beneath you.

> FORWARD 100
The turtle moves forward. A red line appears on floor.svg.

> RIGHT 90
> FORWARD 50
You're near the door-north.

> ENTER door-north
You jump through the door INTO the next room.
The turtle comes with you.
# turtle.yml — a vehicle within room-space
turtle:
  position: {x: 100, y: 100}
  heading: 90  # degrees, 0 = north
  pen_down: true
  pen_color: "#e94560"
  rider: "the-explorer"
CommandEffect
RIDE turtleMount the turtle, move with it
FORWARD 50Move forward, draw if pen down
RIGHT 90Turn right
ENTER doorJump through door to connected room
INTO subroomDescend into nested sub-room
ZOOM OUTSee the room graph navigator

Lineage: Papert's Logo turtle, Rocky's Boots (1982), Robot Odyssey (1984).

Snap Cursor & Pie Menus

When you approach an object, the cursor snaps to it and shows a pie menu of scored actions:

        EXAMINE (80)
           ╱
 REPAIR ──●── USE (95) ← default
           ╲
        TAKE (20)

This IS The Sims interaction model:

  • Objects advertise their available actions
  • Actions are scored based on context, needs, state
  • High-scoring actions appear prominently

Lineage: Don Hopkins' Pie Menus + Will Wright's SimAntics.

Cursor as Vehicle: Direct Manipulation

The cursor carries tools and applies them to the room floor:

> SELECT pen-tool
Cursor now carries: 🖊️ pen (red)

> CLICK workbench
*snap* Cursor at workbench. Pen ready.

> DRAG to door-north
Drawing line from workbench to door-north...
Line added to floor.svg
ToolIconAction
pen🖊️Draw lines on floor
eraser🧽Remove drawings
selector👆Pick up and move objects
linker🔗Draw connections between objects
stamper📌Place copies of cards

Throwing Objects: Data Flow Programming

Throw objects through exits. They pile up on the other side.

> THROW blueprint door-north
Throwing blueprint through door-north...
blueprint landed in assembly/inbox/

Inbox / Outbox

room/
  inbox/           # Objects thrown INTO this room land here
    task-001.yml
  outbox/          # Stage objects before throwing OUT
    result-001.yml
CommandEffect
THROW obj exitToss object through exit
INBOXList waiting items
NEXTProcess next item (FIFO)
STAGE obj exitAdd to outbox
FLUSHThrow all staged objects

Rooms as Pipeline Stages

Each room is a processing node. Exits are edges. Thrown objects are messages.

# Document processing pipeline:
uploads/          # Raw files land here
  inbox/
parser/           # Extract text
  script: parse.py
analyzer/         # LLM analyzes
  prompt: "Summarize and extract entities"
output/           # Final results collect here

This is Kilroy-style data flow: rooms as nodes, files as messages, the filesystem as the network.

Inventories

Characters carry inventories — portable rooms always with them.

# character/inventory/
sword.card
map.yml
notes/
  finding-001.md
CommandEffect
GET swordPick up from room → inventory
DROP mapPut from inventory → room
GIVE torch TO companionTransfer to another character
INVENTList what you're carrying

Your inventory IS a pocket dimension.

Nested Containers

Objects can contain other objects, to arbitrary depth:

> PUT screwdriver IN toolbox
> PUT toolbox IN backpack
> OPEN backpack
backpack contains:
  - toolbox (3 items)
  - sandwich

Object Paths

Address nested objects with paths:

> EXAMINE backpack/toolbox/wrench
> USE inventory/potions/healing
> TAKE ../chest/gold FROM here

Path syntax:

  • container/sub/item — absolute within scope
  • ./toolbox/wrench — relative to current
  • ../sibling/item — parent's sibling
  • /repo-name/path — multi-repo addressing

Tags for Search

> TAG wrench @favorite
> SEARCH backpack @tool
Found in backpack:
  toolbox/screwdriver [@tool]
  toolbox/wrench [@tool @favorite]

Room Graph Navigator

ZOOM OUT to see the whole world:

> ZOOM OUT
│  ROOM GRAPH: moollm-palace              │
│       [room] [card] [chat]              │
│         │                               │
│      [★ YOU ARE HERE]                   │

> LINK room TO card
Connection created. You can now JUMP directly.
CommandEffect
ZOOM OUTSee room graph overview
ZOOM IN roomEnter selected room
LINK a TO bCreate connection between rooms

Like Rocky's Boots: Navigate the structure. Edit while exploring.

Speed of Light vs Carrier Pigeons

Traditional multi-agent: Each agent in isolation. One LLM call per agent. Communication by carrier pigeon. Slow. Expensive. Sad.
MOOLLM: Simulate as many agents together as possible in ONE LLM call. Communication at the speed of light. Multiple simulation steps per iteration.
# In one LLM iteration:
simulation:
  - step: 1
    papert-001: "Microworlds need low floors"
    kay-001: "Yes! Like Smalltalk for children"

  - step: 2
    papert-001:
      responds_to: kay-001
      says: "Exactly! Accessible entry, unlimited ceiling"

  - step: 3
    synthesis:
      emerged: "Low floor + high ceiling + prototypes = MOOLLM"

Three characters, three steps, instant cross-talk — ONE LLM call.

This IS The Sims

The Sims: One frame, all Sims simulated, instant interaction
MOOLLM:   One call, all cards simulated, instant messaging

Instead of isolated agent prisons, we have a shared microworld.

Room Navigation

ActionWhat Happens
EnterPush room's working_set to context
ExitPop context, return to parent
LookRead ROOM.yml and README.md
Activate cardClone card template into room
Complete cardCard writes return_value, can be removed

Nested Rooms (Virtual Zones)

Rooms can contain rooms (subdirectories) or virtual zones (no physical directory):

# cat-cave.yml — TARDIS-like nested room
id: cat-cave
type: [room, furniture]  # Both!

zones:  # Virtual sub-rooms
  nap-zone:
    description: "Sunny spot, cushions everywhere"
    path: "pub/cat-cave/nap-zone"  # Virtual path
  box-jungle:
    description: "Cardboard paradise"
    path: "pub/cat-cave/box-jungle"

Characters reference virtual zones:

# cat-terpie.yml
home: pub/cat-cave/
location: pub/cat-cave/nap-zone  # Virtual zone

Room Relationships

Rooms can remember visitors:

relationships:
  don-hopkins:
    visits: 3
    reputation: "regular"
    memory: "Always nice to the cats"

Home vs Location Protocol

Entities have home (where file lives) and location (where they are):

character:
  home: pub/cat-cave/terpie.yml     # File never moves
  location: pub/                     # Currently in pub

Movement updates location, not file. See character/.

Pie Menu Room Topology

Full specification: TOPOLOGY.yml

The eight-direction compass maps to two types of connections:

DirectionTypeFunction
N/S/E/WCardinalNavigate to major rooms (spiderweb)
NW/NE/SW/SEDiagonalExpand into storage grids (quadrants)

4 ways OUT (navigation) + 4 quadrants IN (infinite storage) = Unlimited worlds

Grid naming: {quadrant}-{x}-{y} (e.g., ne-2-3 = 2 east, 3 north in NE)

See: exit skill for PIE-MENU-TOPOLOGY protocol, memory-palace/ for Method of Loci

Codebase as Navigable World

Directories are rooms. Files are objects. Functions are chambers you can enter.

# Location paths with line numbers
location: "@central/apps/insights/pyleela/brain/Schema.py:142"

# Path syntax
- @repo/path/to/file.py       # file in repo
- @repo/path/to/file.py:42    # specific line
- @repo/path/dir/             # directory (room)

See character/ for party-based code review, README.md for detailed examples.

NPC Embedding Patterns

PatternWhenExample
cat-name.ymlEmbedded NPCpub/cat-terpie.yml
name/CHARACTER.ymlComplex charactercharacters/don-hopkins/
staff-name.ymlRole-based groupingpub/staff-marieke.yml

See naming/ for conventions.

Rooms ARE Logistic Containers!

The Unification: Sims + Factorio

Rooms can participate in a logistics network:

FeatureSourceMOOLLM
Action advertisementsThe SimsObjects advertise what they DO
Item requestsFactorioContainers advertise what they NEED
Attractiveness scoresBothHigher score = higher priority

Room Logistics Mode

room:
  name: "The Kitchen"

  logistics:
    mode: requester              # passive-provider, requester, buffer...
    request_list:
      - tags: ["ingredient"]
        count: 10
        priority: high

Logistic Advertisements

Rooms advertise their NEEDS with attractiveness scores:

logistic_advertisements:

  NEED_INGREDIENTS:
    description: "Kitchen needs ingredients"
    wants:
      - tags: ["ingredient"]
        count: 10
    score: 70                    # Base priority
    score_if: "chef_is_cooking"  # When to boost
    score_bonus: 30              # Total 100 when cooking!

  DESPERATELY_NEED_LIGHT:
    wants:
      - tags: ["light-source"]
        count: 1
    score: 100                   # Very high!

Stacking in Rooms

Rooms can have stacks of items:

room:
  name: "Armory"

  stacks:
    # Fungible (just count)
    arrow: 500
    bolt: 300

    # Instance (individual state)
    magic-sword:
      count: 3
      instances:
        - { id: flame-blade, damage: 50 }
        - { id: frost-edge, damage: 45 }

  stack_limit: 1000

Grid Room Cells

Warehouse rooms can be grid cells:

room:
  name: "Iron Ore Storage"
  type: grid-cell

  grid_cell:
    enabled: true
    parent: "../"
    coordinates: { x: 2, y: 3 }
    item_type: iron-ore

  stacks:
    iron-ore: 2500

The Flow

1. ROOMS advertise logistic needs with scores
2. LOGISTICS ENGINE collects all advertisements
3. Items route to HIGHEST-SCORING requester
4. BOTS or BELTS move items physically
5. ROOM receives items, fires on_item_added

See logistic-container/ and factorio-logistics-protocol.md.


The Philosophy

Spatial navigation IS cognitive navigation.

When you "enter" the debug-session room:

  • Your context shifts to debugging
  • Relevant cards are already in play
  • The room's knowledge is loaded
  • You know where the exits lead

Live Examples

Dovetails With

Sister Skills

Kernel

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

33.85%
按下载量换算34

Claude

27.9%
按下载量换算28

Cursor

19.26%
按下载量换算19

Gemini CLI

9.37%
按下载量换算9

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills