Token导航 LogoToken导航

Agent Skill · 研究检索

empathic-templates(移情模板)

empathic-templates 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。

empathic-templates 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为需要联网;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 66,030。

总安装 269周安装 11GitHub Stars 37下载 87核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

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:empathic-templates(移情模板)
来源仓库:https://github.com/simhacker/moollm
仓库路径:skills/empathic-templates
安装命令:
npx skills add https://github.com/simhacker/moollm --skill empathic-templates
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 empathic-templates

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

empathic-templates 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候选结果。语义感知的模板实例化工具。理解上下文语义而非机械替换。适用于 Handlebars、Jinja 等模板引擎。

适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/empathic-templates

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/simhacker/moollm --skill empathic-templates;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

empathic-templates 怎么安装?

在终端执行 npx skills add https://github.com/simhacker/moollm --skill empathic-templates。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

empathic-templates 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

empathic-templates 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

empathic-templates 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

empathic-templates 用的人多吗?

公开数据是累计安装 269 次,近一周安装 11 次,下载 87 次,GitHub 37 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

empathic-templates 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
empathic-templates

移情模板 · 本页

26937需要联网未明确有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Empathic Templates

*"Templates that understand what you mean, not just what you wrote."*

What Is It?

Empathic Templates are MOOLLM's approach to template instantiation: templates that leverage LLM comprehension to understand semantic intent, not just perform mechanical string substitution.

Traditional templates: {{name}} → replace with literal value Empathic templates: {{name}} → understand what name means in context, generate appropriate content


The Difference

Traditional Templates (Handlebars, Jinja, etc.)

Hello {{name}}!

Your order {{order_id}} has shipped.
Items: {{items}}
Total: ${{total}}

Problem: If items is a list, you need explicit loop syntax. If total should be formatted, you need filters. The template is dumb.

Empathic Templates

# The template understands context
greeting: "{{appropriate_greeting}}"
order_summary: "{{summarize_order_naturally}}"
items: "{{list_items_with_quantities}}"
total: "{{format_currency(total)}}"

The LLM doesn't just substitute — it interprets:

  • {{appropriate_greeting}} becomes "Good morning, Sarah!" or "Hey there!" based on context
  • {{summarize_order_naturally}} becomes prose, not data dump
  • {{list_items_with_quantities}} formats intelligently

How It Works

1. Parse Template

template: |
  # {{character_name}}

  {{describe_appearance}}

  ## Personality
  {{personality_as_prose}}

  ## Catchphrase
  > "{{generate_catchphrase}}"

2. Understand Context

context:
  character_type: "grumpy wizard"
  setting: "medieval fantasy"
  tone: "humorous"
  age: "ancient"

3. Semantic Generation

The LLM doesn't just fill slots — it creates coherent content:

# Grumbald the Insufferable

A stooped figure wrapped in robes that were probably purple
once, before centuries of potion stains and cat hair took
their toll. His beard reaches his knees and contains at
least three quills, a small bird's nest, and what might be
a sandwich from the Third Age.

## Personality

Grumbald has perfected the art of making everyone feel
slightly stupid. He answers questions with questions,
sighs dramatically at simple requests, and has been
known to turn people into newts for using incorrect
grammar. Deep down, he cares — but he'd rather be
turned into a newt himself than admit it.

## Catchphrase

> "I've forgotten more magic than you'll ever know.
> Literally. Where did I put my staff?"

Template Syntax

Basic Slots

{{variable}}           # Simple substitution (still works)
{{expression}}         # Empathic expression (interpreted)
{{#if condition}}      # Conditional (condition is interpreted)
{{#each items}}        # Iteration (smart about structure)
{{> partial}}          # Include another template

Empathic Slots (The Magic)

{{describe_X}}         # Generate description based on context
{{summarize_Y}}        # Create summary of data
{{generate_Z}}         # Create new content fitting context
{{appropriate_W}}      # Choose fitting value from possibilities
{{format_V}}           # Intelligent formatting

Expressions with Empathic Interpretation

{{#if character.is_hungry}}
  {{describe_hunger_behavior}}
{{/if}}

{{#if player.reputation > 50}}
  # The guard recognizes you
  {{guard_friendly_greeting}}
{{else}}
  # The guard is suspicious
  {{guard_suspicious_challenge}}
{{/if}}

The conditions use Empathic Expressions for flexible interpretation.


Used For

Character Creation

# CHARACTER.yml.tmpl
id: {{generate_unique_id}}
name: "{{character_name}}"
species: "{{species}}"
description: |
  {{describe_appearance_based_on_species_and_personality}}

personality:
  {{infer_sims_traits_from_description}}

catchphrase: "{{generate_fitting_catchphrase}}"

backstory: |
  {{generate_backstory_consistent_with_setting}}

Room Generation

# ROOM.yml.tmpl
room:
  name: "{{room_name}}"
  type: {{room_type}}

  description: |
    {{describe_room_atmospherically}}

    {{describe_notable_features}}

    {{hint_at_secrets_if_any}}

  exits:
    {{generate_exits_based_on_maze_topology}}

  objects:
    {{place_appropriate_objects}}

Session Logs

# SESSION.md.tmpl
## Session {{session_number}} — {{session_date}}

### Summary
{{summarize_session_events}}

### Key Moments
{{list_memorable_moments}}

### State Changes
{{document_world_changes}}

### Next Steps
{{suggest_continuation_hooks}}

The Empathic Expression Connection

Templates use Empathic Expressions for:

Variables

{{user.name}}                    # Simple lookup
{{user.status | capitalize}}     # With filter
{{player.has_item('key')}}       # Method call

Conditions

{{#if character.mood == 'happy'}}  # Exact match
{{#if character.is_friendly}}       # Boolean inference
{{#if gold > 100}}                  # Numeric comparison
{{#if player.can_afford(item)}}     # Method evaluation

Iterations

{{#each inventory.items}}
  - {{this.name}}: {{this.description}}
{{/each}}

Code-Switching IN Templates

# Generate SQL for the report
query: |
  {{empathic_sql: "get all users who ordered in the last month"}}

# Generate the email
email: |
  Dear {{user.name}},

  {{summarize_monthly_orders_naturally}}

  Total spent: {{format_currency(monthly_total)}}

Template Discovery Pattern

First 50 lines sniff:

Templates should front-load:

  1. Template metadata (name, purpose, required context)
  2. Required variables list
  3. Optional variables with defaults
# CHARACTER.yml.tmpl — Character sheet template
#
# Required context:
#   - character_name: string
#   - species: string
#   - setting: string (e.g., "medieval fantasy")
#
# Optional context:
#   - personality_hints: list of traits
#   - backstory_seeds: key events to include
#   - tone: "serious" | "humorous" | "dark" (default: "neutral")

id: {{generate_unique_id}}
name: "{{character_name}}"
# ... rest of template

LLM can sniff first 50 lines to understand what the template needs before reading the full file.


Comment Intelligence

The LLM distinguishes between meta-comments (instructions for generation) and concrete comments (meant for output):

Meta-Comments (Stripped)

# TEMPLATE: This section describes the character's appearance
# INSTRUCTION: Use vivid sensory details
# NOTE: Keep under 100 words
# TODO: Add more variety to hair colors
description: |
  {{describe_appearance}}

These are instructions TO the LLM. They guide generation but don't appear in output.

Concrete Comments (Preserved)

# This character was created using the incarnation protocol.
# See skills/incarnation/SKILL.md for details.
description: |
  {{describe_appearance}}

# Sims traits determine interaction success rates
sims_traits:
  {{generate_traits}}

These are comments FOR the output file. They explain the generated content to future readers.

How the LLM Knows

IndicatorTypeAction
# TEMPLATE:, # INSTRUCTION:, # NOTE:MetaStrip
# TODO:, # FIXME: in template contextMetaStrip
# This explains..., # See also...ConcretePreserve
Comments inside {{...}} blocksMetaStrip
Comments explaining generated valuesConcretePreserve
ALL CAPS directive styleMetaStrip
Lowercase explanatory styleConcretePreserve

Example: Mixed Comments

Template:

# TEMPLATE: Character soul file
# INSTRUCTION: Generate YAML Jazz style comments

# This character was incarnated via the full autonomy protocol.
id: {{generate_id}}
name: "{{character_name}}"

# INSTRUCTION: Describe based on species and personality
description: |
  {{describe_appearance}}

# Personality traits affect all interactions
# Higher values = stronger tendency
sims_traits:
  nice: {{nice_value}}  # 0-10, affects social success
  # INSTRUCTION: Infer from description
  playful: {{infer_playful}}

Generated Output:

# This character was incarnated via the full autonomy protocol.
id: palm-001
name: "Palm"

description: |
  A small capuchin monkey with knowing eyes and surprisingly
  dexterous fingers. Palm moves with the confidence of someone
  who has already seen how this story ends.

# Personality traits affect all interactions
# Higher values = stronger tendency
sims_traits:
  nice: 7  # 0-10, affects social success
  playful: 9

What happened:

  • # TEMPLATE: and # INSTRUCTION: lines → stripped
  • # This character was incarnated... → preserved
  • # Personality traits affect... → preserved
  • # 0-10, affects social success → preserved
  • # INSTRUCTION: Infer from description → stripped

The Principle

Meta-comments teach the generator. Concrete comments teach the reader.

The LLM understands this distinction because it understands intent. Directive language instructs; explanatory language documents.


Relationship to Self-Style Inheritance

Templates are prototypes. Instantiation creates instances:

prototype: skills/character/CHARACTER.yml.tmpl
  ↓ instantiate with context
instance: examples/adventure-4/characters/animals/monkey-palm/CHARACTER.yml

The template defines shape and intent. The instance contains specific values.

But empathic templates go further: they generate appropriate content, not just fill slots.


Examples

Instantiating a Buff

# BUFF.yml.tmpl
buff:
  id: {{buff_id}}
  name: "{{buff_name}}"
  type: {{buff_type}}

  description: |
    {{describe_buff_effect_flavorfully}}

  mechanics:
    duration: {{duration}}
    effect: "{{describe_mechanical_effect}}"

  trigger: "{{empathic_expression: trigger_condition}}"

  flavor_text: |
    {{generate_flavor_text_matching_setting}}

Context:

buff_name: "Sugar Rush"
buff_type: positive
trigger_condition: "eating candy or sweet treats"
setting: "whimsical candy land"

Generated:

buff:
  id: sugar-rush
  name: "Sugar Rush"
  type: positive

  description: |
    Your eyes widen, your heart races, and suddenly
    EVERYTHING IS AMAZING. Colors are brighter, sounds
    are sweeter, and you're pretty sure you could
    outrun a unicorn.

  mechanics:
    duration: 10 minutes
    effect: "+20% speed, +10% charisma, -10% focus"

  trigger: "consuming_sweet_treat"

  flavor_text: |
    *The sugar hits your bloodstream like a
    candy-coated lightning bolt. WHEEEEE!*

Templates as Schemas (CRITICAL!)

Templates are not just for instantiation — they ARE THE SCHEMA. The same .tmpl file serves as:

  1. Human documentation — What fields exist, what they mean
  2. Machine schema — Required vs optional, types, constraints
  3. Prototype definition — Default values, inherited structure
  4. Code generation source — Python dataclasses, JS classes, validators

The Template-Schema Pattern

# ADVENTURE.yml.tmpl — Adventure State Schema
#
# REQUIRED fields (must be provided):
#   - adventure.name: string
#   - player.character: path
#   - navigation.starting_room: path
#
# OPTIONAL fields (can be omitted, inherited, or use defaults):
#   - parameters.*: all have sensible defaults
#   - party.members: defaults to [player.character]
#   - evidence.*: starts empty
#
# COMPUTED fields (LLM generates):
#   - adventure.started: timestamp
#   - selection.targets: starts []

Smart Instantiation: The Drop Pattern

Traditional: Fill every slot, even if the value is default.

Empathic: DROP optional sections entirely if:

  • The value equals the prototype default
  • The context doesn't need it
  • An abstract description is sufficient
# Template has:
parameters:
  time:
    advancement: normal      # Default
  git:
    auto_commit: false       # Default
    auto_push: false         # Default
  debug:
    enabled: false           # Default
    # ...50 more lines of debug config...

# Smart instantiation DROPS the whole section:
# (No parameters block = use all defaults)

# Unless something differs:
parameters:
  debug:
    enabled: true            # Only what changed!

Schema Markers in Templates

Use special comments to mark field requirements:

# REQUIRED: Must be provided during instantiation
adventure:
  name: "{{adventure_name}}"           # REQUIRED
  objective: "{{quest_objective}}"     # REQUIRED
  status: active                       # OPTIONAL: has default

# OPTIONAL: Can be omitted entirely
evidence:                              # OPTIONAL: section
  clues: []                            # OPTIONAL: default []
  items: []                            # OPTIONAL: default []

# COMPUTED: Generated by LLM or system
statistics:                            # COMPUTED: section
  rooms_explored: 0                    # COMPUTED: auto-increment
  turns_elapsed: 0                     # COMPUTED: auto-increment

# INHERITED: Comes from prototype
parameters:                            # INHERITED: from simulation skill
  # Only override what differs from skills/simulation/defaults.yml

Field Requirement Markers

MarkerMeaningInstantiation Behavior
# REQUIREDMust be filledError if missing
# OPTIONALCan be omittedDrop if default or empty
# OPTIONAL: default XHas default valueDrop if equals X
# COMPUTEDSystem generatesNever fill from context
# INHERITEDFrom prototypeDrop if unchanged
# ABSTRACTNatural language OKKeep as prose placeholder

Abstract Fields: Natural Language as Value

Sometimes the "value" is just a description of intent:

# Template:
room:
  atmosphere: "{{atmosphere}}"         # ABSTRACT: describe the feeling

# Instantiation with abstract value (valid!):
room:
  atmosphere: |
    A sense of ancient mystery. Dust motes float in shafts of light
    from high windows. Something important happened here, long ago.

# The LLM can work with this! It doesn't need structured data.

Prototype Inheritance

Templates inherit from prototypes. Only override what differs:

# skills/room/ROOM.yml.tmpl is the PROTOTYPE
# examples/adventure-4/maze/ROOM.yml is an INSTANCE

# The instance OMITS fields that match the prototype:
room:
  name: "Maze Entrance"     # DIFFERENT: specific name
  # purpose: omitted        # INHERITED: from prototype
  # working_set: omitted    # INHERITED: from prototype
  exits:
    north: corridor-1/      # DIFFERENT: specific exits
    east: dead-end/
  atmosphere: "Confusion and possibility"  # DIFFERENT: specific vibe

Code Generation from Templates

Templates inform Python and JavaScript class generation:

# adventure/schema.py — GENERATED FROM ADVENTURE.yml.tmpl

@dataclass
class Adventure:
    # REQUIRED fields (from template markers)
    name: str
    objective: str
    starting_room: str
    player_character: str

    # OPTIONAL fields (have defaults)
    status: str = "active"
    parameters: Optional[Parameters] = None  # Whole section optional

    # COMPUTED fields (not in __init__)
    started: str = field(init=False)
    turns_elapsed: int = field(init=False, default=0)

    # Unknown fields preserved for round-trip
    _extra: Dict[str, Any] = field(default_factory=dict)

    def __post_init__(self):
        self.started = datetime.now().isoformat()
// engine.js — GENERATED FROM templates

class Adventure {
  // REQUIRED
  name;        // string, must be set
  objective;   // string, must be set

  // OPTIONAL with defaults
  status = "active";
  parameters = null;  // null = use global defaults

  // COMPUTED (read-only, managed by engine)
  get started() { return this._started; }
  get turns_elapsed() { return this._turns; }

  constructor(data) {
    // Validate REQUIRED
    if (!data.name) throw new Error("Adventure requires name");
    if (!data.objective) throw new Error("Adventure requires objective");

    // Apply provided values
    Object.assign(this, data);

    // Preserve unknown fields
    this._extra = {};
    for (const [k, v] of Object.entries(data)) {
      if (!this.constructor.KNOWN_FIELDS.includes(k)) {
        this._extra[k] = v;
      }
    }
  }
}

LLM Compilation Events

When the template contains expressions, emit events for the LLM to compile:

# Template with expressions:
guard:
  allows_entry: "{{empathic_expression: player has the key OR player is known to guard}}"
  greeting: "{{generate: appropriate greeting based on player reputation}}"
  actions:
    - trigger: "{{when: player attempts to pass without permission}}"
      action: "{{do: block and challenge}}"
      score: "{{calculate: based on guard alertness and player stealth}}"

The adventure.py linter emits events:

# Events for LLM compilation:
- event: COMPILE_EXPRESSION
  field: "guard.allows_entry"
  source: "player has the key OR player is known to guard"
  target_language: javascript
  expected_type: boolean
  output_field: "guard.allows_entry_js"  # Where to write the compiled expression
  # Naming convention: {field}_js for JavaScript, {field}_py for Python
  # The corresponding runtime classes know to eval these fields

- event: COMPILE_GENERATION
  field: "guard.greeting"
  instruction: "appropriate greeting based on player reputation"
  target_format: string
  context_needed: [player.reputation, guard.personality]

- event: COMPILE_SCORE
  field: "guard.actions[0].score"
  instruction: "based on guard alertness and player stealth"
  target_format: "number 0-100"
  inputs: [guard.alertness, player.stealth]

The LLM responds with compiled expressions, written to the _js or _py suffixed fields:

# After LLM compilation, the YAML now contains:
guard:
  allows_entry: "player has the key OR player is known to guard"
  allows_entry_js: |
    (ctx) => ctx.player.inventory.includes('key') ||
             ctx.guard.knownPlayers.includes(ctx.player.id)
  allows_entry_py: |
    lambda ctx: 'key' in ctx.player.inventory or
                ctx.player.id in ctx.guard.known_players

  greeting: "appropriate greeting based on player reputation"
  greeting_js: |
    (ctx) => {
      if (ctx.player.reputation > 80) return "Welcome back, friend!";
      if (ctx.player.reputation > 50) return "You may pass.";
      return "Halt! State your business.";
    }

Output Field Naming Convention

FieldOutput Field (JS)Output Field (Py)
allows_entryallows_entry_jsallows_entry_py
scorescore_jsscore_py
guard.conditionguard.condition_jsguard.condition_py

The runtime classes know to look for these suffixed fields:

// engine.js runtime
class Guard {
  checkEntry(ctx) {
    if (this.allows_entry_js) {
      return eval(this.allows_entry_js)(ctx);  // Use compiled
    }
    return this.evaluateEmpathic(this.allows_entry, ctx);  // LLM fallback
  }
}
# schema.py runtime
class Guard:
    def check_entry(self, ctx):
        if self.allows_entry_py:
            return eval(self.allows_entry_py)(ctx)  # Use compiled
        return self.evaluate_empathic(self.allows_entry, ctx)  # LLM fallback

This creates a graceful degradation pattern:

  1. If _js or _py exists → use fast compiled expression
  2. If not → fall back to LLM interpretation of natural language

The same YAML file works in both compiled (fast) and interpreted (flexible) modes!

The Full Template-to-Code Pipeline

# Template-to-Code Pipeline
template_pipeline:
  - stage: "ADVENTURE.yml.tmpl"
    role: "Human-readable template/schema"
  - stage: "adventure.py lint"
    role: "Parse, validate, emit events"
    actions:
      - "Validate REQUIRED fields"
      - "Warn on missing OPTIONAL with no default"
      - "Emit COMPILE_EXPRESSION events"
      - "Emit COMPILE_GENERATION events"
      - "Emit COMPILE_SCORE events"
  - stage: "LLM receives events"
    role: "Compiles expressions"
  - stage: "adventure.py compile"
    role: "Generate output"
    outputs:
      - "adventure.json (minimal, machine-readable)"
      - "engine.js (with compiled expressions)"
      - "schema.py (Python classes for validation)"
      - "index.html (playable web app)"

Validation and Linting

Templates enable automated validation:

# Linter checks:
validation:
  required_present:
    - adventure.name
    - adventure.objective
    - player.character
    - navigation.starting_room

  type_checks:
    - field: simulation.turn
      type: integer
      min: 0

    - field: parameters.difficulty
      type: enum
      values: [easy, normal, hard]

  expression_checks:
    - field: guard.allows_entry
      must_compile_to: boolean

  path_checks:
    - field: player.character
      must_exist: true
      must_be_type: directory

Example: Smart Instantiation

Template (ROOM.yml.tmpl):

room:
  name: "{{name}}"                    # REQUIRED
  purpose: "{{purpose}}"              # OPTIONAL: default "general"
  created: "{{timestamp}}"            # COMPUTED

  context: []                         # OPTIONAL: default []
  cards_in_play: []                   # OPTIONAL: default []

  working_set:                        # INHERITED: from prototype
    - "ROOM.yml"
    - "README.md"
    - "state/*.yml"

  exits:                              # REQUIRED: at least one
    parent: "../"

  atmosphere: "{{atmosphere}}"        # ABSTRACT: natural language OK

Minimal valid instantiation:

# Only what's needed and different:
room:
  name: "Wizard's Study"
  exits:
    parent: "../"
    down: secret-lab/
  atmosphere: "Dusty tomes and arcane instruments"

What the linter infers:

  • ✅ name provided (REQUIRED)
  • ✅ exits has at least one (REQUIRED)
  • ⚡ purpose missing → use default "general"
  • ⚡ context missing → use default []
  • ⚡ cards_in_play missing → use default []
  • ⚡ working_set missing → inherit from prototype
  • ⚡ created missing → compute timestamp
  • ✅ atmosphere is natural language (ABSTRACT valid)

Anti-Pattern: Dumb Templates

Don't do this:

description: "{{description}}"  # Just passes through
personality: "{{personality}}"  # No generation

Do this:

description: |
  {{describe_character_based_on_traits_and_setting}}

personality:
  {{infer_traits_from_context}}

Let the LLM add value. That's the whole point.


Anti-Pattern: Repeating Defaults

Don't do this:

# Every room file repeats the same working_set:
working_set:
  - "ROOM.yml"
  - "README.md"
  - "state/*.yml"

Do this:

# Omit working_set entirely — inherit from ROOM.yml.tmpl prototype
# Only include if this room has DIFFERENT requirements

Templates enable inheritance. Use it!


Dovetails With


Protocol Symbol

EMPATHIC-TEMPLATES

Invoke when: Instantiating templates with semantic understanding.

See: PROTOCOLS.yml

平台分布

Codex

33.12%
按下载量换算29

Claude

31.69%
按下载量换算28

Cursor

17.44%
按下载量换算15

Gemini CLI

9.87%
按下载量换算9

安全审计

Gen Agent Trust Hub

可疑

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills