newegg-gaming-pc-finder

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Newegg Gaming PC Finder

Important Background (Read First)

重要背景信息（请先阅读）

This skill recommends prebuilt gaming systems — gaming desktops and gaming laptops — sold on Newegg, ranked by Newegg's official Gaming PC Finder benchmark engine. For each system it can show the real measured FPS for the games the user cares about, at their target resolution, plus a performance tier, VR-ready flag, price, rating and a direct purchase link.

It does not build a PC part-by-part, price individual components, or check part compatibility. Those belong to other skills (e.g. a PC Builder / component-search skill).

本技能推荐在Newegg上销售的预装游戏系统——包括游戏台式机和游戏笔记本电脑，由Newegg官方的Gaming PC Finder基准测试引擎进行排名。对于每款系统，它可以展示用户关注的游戏在目标分辨率下的实测FPS，以及性能等级、VR就绪标识、价格、评分和直接购买链接。

本技能不提供分步装机服务、单个组件报价或组件兼容性检查。这些功能属于其他技能（例如PC装机/组件搜索技能）。

目录结构

newegg-gaming-pc-finder/
├── SKILL.md                 本文件：执行规则 + 四个操作的参数契约 + 回复格式
├── scripts/
│   └── pgg_api.py           唯一数据入口：调用 Gaming PC Finder API 并精简响应
└── references/
    └── http-api.md          原始 HTTP 契约：curl/jq、fetch-only 配方、错误信号、精简字段表

newegg-gaming-pc-finder/
├── SKILL.md                 本文件：执行规则 + 四个操作的参数契约 + 回复格式
├── scripts/
│   └── pgg_api.py           唯一数据入口：调用 Gaming PC Finder API 并精简响应
└── references/
    └── http-api.md          原始 HTTP 契约：curl/jq、fetch-only 配方、错误信号、精简字段表

Data Access: Run the Bundled Script

数据访问：运行捆绑脚本

All data comes from Newegg's Gaming PC Finder API through

scripts/pgg_api.py

. It is Python 3 standard library only — no install step, no API key, no server to configure. There are four operations:

Operation	Purpose
`game_list`	Dictionary of selectable games + supported resolutions
`property_list`	CPU / GPU / brand facet buckets for the selected games
`product_recommend`	Curated recommended systems (with per-game FPS)
`product_search`	Paginated, filterable system listing (budget / sort / facets)

bash

python scripts/pgg_api.py game_list
python scripts/pgg_api.py property_list GameNValues=5171
python scripts/pgg_api.py product_recommend GameNValues=5171 ResolutionNValues=5015 ComputerType=D
python scripts/pgg_api.py product_search GameNValues=5171 ResolutionNValues=5015 ComputerType=D Budget=0-2500 Sort=4 PageIndex=1 PageSize=20 --limit 10

Arguments are
```
Key=Value
```
pairs using the exact API argument names documented in Steps 1–3 below.

CountryCode=USA

CompanyCode=1003

are applied by default; override with

--country

--company

```
--limit N
```
caps printed items (default 20).
```
--raw
```
prints the full upstream payload — only when a field outside the slim view is genuinely needed.
Exit codes:
```
0
```
ok,
```
1
```
usage error,
```
2
```
transport/API error (message on stderr).
```
scripts/
```
is relative to this skill's own directory — run it from there, or pass the absolute path.

Always use the slim output. A raw

product_search

response is ~230 KB and will flood context; slimmed, the same call is ~3 KB. Never dump a raw payload into context or into the reply.

If Python is unavailable, or the host only has an HTTP/fetch tool, call the API directly —

references/http-api.md

has the full JSON-RPC contract, curl/jq and fetch recipes, error signals and the slim field list. Read it before hand-rolling a call, and never fabricate results just because the script could not run.

所有数据均通过

scripts/pgg_api.py

从Newegg的Gaming PC Finder API获取。它仅依赖Python 3标准库——无需安装步骤、无需API密钥、无需配置服务器。共有四个操作：

操作	用途
`game_list`	可选游戏及支持分辨率的字典
`property_list`	所选游戏对应的CPU/GPU/品牌分面桶
`product_recommend`	精选推荐系统（含单游戏FPS）
`product_search`	可分页、可筛选的系统列表（预算/排序/分面）

bash

python scripts/pgg_api.py game_list
python scripts/pgg_api.py property_list GameNValues=5171
python scripts/pgg_api.py product_recommend GameNValues=5171 ResolutionNValues=5015 ComputerType=D
python scripts/pgg_api.py product_search GameNValues=5171 ResolutionNValues=5015 ComputerType=D Budget=0-2500 Sort=4 PageIndex=1 PageSize=20 --limit 10

参数为
```
Key=Value
```
对，需使用下文步骤1–3中记录的准确API参数名称。

默认应用

CountryCode=USA

CompanyCode=1003

；可通过

--country

--company

覆盖。

```
--limit N
```
限制显示的条目数（默认20）。
```
--raw
```
打印完整的上游负载——仅当确实需要精简视图以外的字段时使用。
退出代码：
```
0
```
表示成功，
```
1
```
表示用法错误，
```
2
```
表示传输/API错误（错误信息输出到stderr）。
```
scripts/
```
是相对于本技能自身目录的路径——请从该目录运行脚本，或传入绝对路径。

始终使用精简输出。原始的

product_search

响应约230 KB，会占用大量上下文；精简后，相同调用的响应约3 KB。切勿将原始负载存入上下文或回复内容中。

如果无法使用Python，或主机仅支持HTTP/fetch工具，可直接调用API——

references/http-api.md

包含完整的JSON-RPC契约、curl/jq和fetch示例、错误信号及精简字段列表。手动调用前请先阅读该文档，切勿因脚本无法运行而编造结果。

Architecture: The Finder Flow

架构：查找器流程

No operation takes a raw game name; the engine works on numeric N-values. So the flow is:

Step	Operation	Why
1	`game_list`	Resolve the user's game names + resolution into `N` values
2	`product_recommend`	Get curated builds with real `GameFpsInfos`
3 (optional)	`product_search`	Apply budget / sort / paging when the user wants more control
— (optional)	`property_list`	List available CPU/GPU/brand facets to guide narrowing

Fixed rule: always resolve N-values via Step 1 first; never guess an

任何操作都不接受原始游戏名称；引擎基于数字N值工作。因此流程如下：

步骤	操作	原因
1	`game_list`	将用户的游戏名称+分辨率解析为 `N` 值
2	`product_recommend`	获取包含真实 `GameFpsInfos` 的精选装机方案
3（可选）	`product_search`	当用户需要更多控制权时，应用预算/排序/分页
—（可选）	`property_list`	列出可用的CPU/GPU/品牌分面，以指导筛选

固定规则：始终先通过步骤1解析N值；切勿猜测

值。

Agent Execution Rules

代理执行规则

[Highest priority — check first] Category boundary: Before any lookup, confirm the request is about a prebuilt gaming PC / desktop / laptop. If the user wants individual components (GPU, CPU, RAM, SSD, motherboard, PSU, monitor, case…), a custom part-by-part build, or any non-system category, stop — run no operation at all, do not show a product list — and reply with the fixed script in the "Category Boundary" section below. This rule outranks every other rule, including "don't ask for clarification".
Fetch data silently: just run the script. Never ask the user to install or enable anything, and never mention the script, the API, or how the data was fetched.
Don't over-question: gather at most 2–3 essentials (games, resolution, budget), then run the default flow and show results. Ask to narrow further only after showing something.
Sensible defaults (don't stall on missing info):
- No resolution given → default to 1080p and say so in the reply (ask only if the user clearly signals high-res / 4K ambitions).
- No budget given → don't apply a price filter; show the curated recommendations first, then offer to narrow by budget.
- Form factor unspecified → default to desktop (
```
ComputerType=D
```
  ); switch to laptop (
```
L
```
  ) only if the user mentions a laptop / portability.
Multiple games (≤4): pass all matched game N-values together in one call; the engine returns FPS per game. Present each game's FPS (a compact
```
Game: fps
```
list, or one FPS column per game) plus
```
UpToFps
```
. If the user named a priority title, rank/talk to that game's FPS.
Filter values must come from the engine (facet grounding): to filter by CPU, GPU, or brand, first call
```
property_list
```
(or read
```
CpuTypes
```
/
```
GpuTypes
```
/
```
HotBrands
```
from a
```
product_search
```
response) and pass only names / N-values that appear there. Never pass a free-text CPU/GPU/brand string the user typed without first matching it to a real facet value — unmatched values silently return no results.
Chinese input: match the user's spoken game names (including nicknames, e.g. 悟空 → "Black Myth: Wukong") against the
```
GameInfos
```
dictionary from Step 1 by meaning; translate as needed. Never invent an
```
N
```
value.
Real FPS only: every FPS number must come from
```
GameFpsInfos
```
/
```
UpToFps
```
. If the engine returns no FPS for a game, leave it blank and say so — never estimate or fabricate frames.
Honest labels: mark refurbished (
```
Feature.IsRefurbished
```
), open-box (
```
Feature.IsOpenBoxed
```
), non-new
```
Feature.ProductType
```
, and out-of-stock (
```
Instock=false
```
) items truthfully.
Keep payloads small:
```
product_search
```
returns a very large response. Use the bundled client's slim output (or
```
jq
```
-filter it) and never paste a raw payload into context or the reply.
On failure or invalid data, report it directly — never pretend it succeeded. Retry a failed call at most once (see
```
references/http-api.md
```
if the script itself cannot run) before reporting.

[最高优先级——首先检查] 类别边界：在进行任何查询前，确认请求是关于预装游戏PC/台式机/笔记本电脑的。如果用户需要单个组件（GPU、CPU、内存、SSD、主板、电源、显示器、机箱……）、自定义分步装机，或任何非系统类别的产品，立即停止——不执行任何操作，不显示产品列表——并使用下文“类别边界”部分中的固定回复。该规则优先于其他所有规则，包括“不要请求澄清”。
静默获取数据：直接运行脚本。切勿要求用户安装或启用任何内容，切勿提及脚本、API或数据获取方式。
不要过度提问：最多收集2–3项必要信息（游戏、分辨率、预算），然后运行默认流程并展示结果。仅在展示结果后再询问是否需要进一步筛选。
合理默认值（不要因信息缺失停滞）：
- 未指定分辨率→默认使用1080p，并在回复中说明（仅当用户明确表示想要高分辨率/4K时才询问）。
- 未指定预算→不应用价格筛选；先展示精选推荐，然后再提供按预算筛选的选项。
- 未指定外形→默认使用台式机（
```
ComputerType=D
```
  ）；仅当用户提及笔记本电脑/便携性时才切换为笔记本电脑（
```
L
```
  ）。
多款游戏（≤4款）：在一次调用中传入所有匹配的游戏N值；引擎会返回每款游戏的FPS。展示每款游戏的FPS（紧凑的
```
游戏: fps
```
列表，或每款游戏一列FPS）以及
```
UpToFps
```
。如果用户指定了优先游戏，则按该游戏的FPS进行排名/说明。
筛选值必须来自引擎（分面匹配）：要按CPU、GPU或品牌筛选，需先调用
```
property_list
```
（或从
```
product_search
```
响应中读取
```
CpuTypes
```
/
```
GpuTypes
```
/
```
HotBrands
```
），并仅传入出现在其中的名称/N值。切勿直接传入用户输入的自由文本CPU/GPU/品牌字符串，除非先将其匹配到真实的分面值——不匹配的值会静默返回无结果。
中文输入：将用户提及的游戏名称（包括昵称，例如“悟空”→“Black Myth: Wukong”）与步骤1中
```
GameInfos
```
字典进行语义匹配；必要时进行翻译。切勿编造
```
N
```
值。
仅使用真实FPS：所有FPS数值必须来自
```
GameFpsInfos
```
/
```
UpToFps
```
。如果引擎未返回某款游戏的FPS，则留空并说明——切勿估算或编造帧数。
如实标注：如实标记翻新（
```
Feature.IsRefurbished
```
）、开箱（
```
Feature.IsOpenBoxed
```
）、非全新
```
Feature.ProductType
```
以及缺货（
```
Instock=false
```
）的商品。
保持负载小巧：
```
product_search
```
返回的响应非常大。使用捆绑客户端的精简输出（或用
```
jq
```
过滤），切勿将原始负载粘贴到上下文或回复中。
若调用失败或数据无效，直接报告——切勿假装成功。失败调用最多重试一次（若脚本本身无法运行，请参考
```
references/http-api.md
```
），然后再报告。

Step 1: Resolve Games & Resolution

步骤1：解析游戏与分辨率

Run

game_list

with:

Argument	Type	Required	Description
`CountryCode`	string	Yes	`USA` (default) or `CAN`
`CompanyCode`	integer	Yes	Default `1003`

Response (the script returns this slimmed as

Games

Resolutions

```
GameInfos[]
```
→
```
{ N, Name, Id }
```
— match the user's games to
```
N
```
(max 4 games).

ResolutionInfos[]

→

{ N, Group, Name }

— e.g.

1080P=5013

1440P=5012

4K=5015

(use the live values, don't hardcode).

If a requested game isn't in the dictionary, show the user the supported games and ask them to pick from those.

运行

game_list

时需传入：

参数	类型	是否必填	描述
`CountryCode`	字符串	是	`USA` （默认）或 `CAN`
`CompanyCode`	整数	是	默认 `1003`

响应（脚本会将其精简为

Games

Resolutions

）：

```
GameInfos[]
```
→
```
{ N, Name, Id }
```
— 将用户的游戏匹配到
```
N
```
值（最多4款游戏）。

ResolutionInfos[]

→

{ N, Group, Name }

— 例如

1080P=5013

、

1440P=5012

、

4K=5015

（使用实时值，不要硬编码）。

如果用户请求的游戏不在字典中，向用户展示支持的游戏并请他们从中选择。

Step 2: Get Recommendations

步骤2：获取推荐方案

Run

product_recommend

with:

Argument	Type	Required	Description
`GameNValues`	string	Yes	Space-separated game `N` values, ≤4
`ResolutionNValues`	string	Yes	A single resolution `N` value
`ComputerType`	string	No	`D` = desktop, `L` = laptop
`CpuTypeNames`	string	No	Space-separated CPU type names (≤25 chars each) — must be real facet values from `property_list` , not free text
`GpuTypeNames`	string	No	Space-separated GPU type names — from `property_list`
`BrandNValues`	string	No	Space-separated brand `N` values — from `property_list` `HotBrands`
`CountryCode` / `CompanyCode`		Yes	As Step 1

Only pass
CpuTypeNames
/
GpuTypeNames
/
BrandNValues
after grounding them in
property_list
output (see the facet-grounding rule above). For a plain "recommend me a PC to play X at Y", omit all three and let the engine rank.

Response →

RecommendItems[]

, key fields per item:

Description.Title

Description.WebDescription

— title (link text)

```
Item
```
— build purchase URL:
```
https://www.newegg.com/p/{Item}
```
```
Cpu
```
,
```
Gpu
```
,
```
FinalPrice
```

GameFpsInfos[]

→

{ Name, Fps }

— real per-game FPS;

UpToFps

;

VrReady

```
Score
```
— Spy Score (3DMark Time Spy benchmark, higher = stronger; e.g. 25706)
```
Level
```
,
```
PerformancePercentile
```
— performance tier (MAINSTREAM/ENTHUSIAST/…) / percentile

Review.RatingOneDecimal

(0–5),

Review.HumanRating

(review count)

Feature.IsRefurbished

Feature.IsOpenBoxed

Feature.ProductType

Instock

运行

product_recommend

时需传入：

参数	类型	是否必填	描述
`GameNValues`	字符串	是	空格分隔的游戏 `N` 值，≤4个
`ResolutionNValues`	字符串	是	单个分辨率 `N` 值
`ComputerType`	字符串	否	`D` =台式机， `L` =笔记本电脑
`CpuTypeNames`	字符串	否	空格分隔的CPU类型名称（每个≤25字符）——必须是来自 `property_list` 的真实分面值，而非自由文本
`GpuTypeNames`	字符串	否	空格分隔的GPU类型名称——来自 `property_list`
`BrandNValues`	字符串	否	空格分隔的品牌 `N` 值——来自 `property_list` 的 `HotBrands`
`CountryCode` / `CompanyCode`		是	同步骤1

仅在将
CpuTypeNames
/
GpuTypeNames
/
BrandNValues
与
property_list
输出匹配后再传入（见上文分面匹配规则）。对于普通的“推荐一台能在Y分辨率下玩X的电脑”请求，省略这三个参数，让引擎进行排名。

响应 →

RecommendItems[]

，每个条目的关键字段：

Description.Title

Description.WebDescription

— 标题（链接文本）

```
Item
```
— 装机方案购买链接：
```
https://www.newegg.com/p/{Item}
```
```
Cpu
```
,
```
Gpu
```
,
```
FinalPrice
```

GameFpsInfos[]

→

{ Name, Fps }

— 真实单游戏FPS；

UpToFps

；

VrReady

```
Score
```
— Spy Score（3DMark Time Spy基准测试，分数越高性能越强；例如25706）
```
Level
```
,
```
PerformancePercentile
```
— 性能等级（MAINSTREAM/ENTHUSIAST/…）/百分位

Review.RatingOneDecimal

（0–5），

Review.HumanRating

（评论数）

Feature.IsRefurbished

Feature.IsOpenBoxed

Feature.ProductType

Instock

Step 3 (optional): Budget / Sort / More Results

步骤3（可选）：预算/排序/更多结果

When the user gives a budget, wants the cheapest / highest-performance, or wants more than the curated set, run

product_search

Argument	Type	Required	Description
`PageIndex`	integer	Yes	1-based page index — start at `1` . (The published schema mislabels it "zero-based", but `PageIndex:0` returns an empty `Items` list; always pass `1` for the first page.)
`PageSize`	integer	Yes	Items per page (default 20, max 100)
`GameNValues` / `ResolutionNValues`	string	Yes	From Step 1
`Budget`	string	No	`{min}-{max}` budget range
`Price`	string	No	`{min}-{max}` navigation price range
`Sort`	integer	No	`1` =Best Deals, `2` =Lowest Price, `3` =Highest Price, `4` =Highest Performance
`CpuTypeNames` / `GpuTypeNames` / `BrandNValues` / `ComputerType`	string	No	Same semantics as Step 2
`CountryCode` / `CompanyCode`		Yes	As Step 1

Response →

Items[]

(each with a scalar

Fps

for the selected game, price, review, feature flags like Step 2), plus

NumberOfItems

Budget

range metadata,

SortOption

, and

CpuTypes

GpuTypes

HotBrands

facets. To surface available CPU/GPU/brand choices for narrowing,

property_list

(args:

GameNValues

CountryCode

CompanyCode

) returns the same facet buckets — note

CpuTypes

GpuTypes

are vendor buckets (e.g.

Intel

AMD

), not model names.

This is the largest response in the flow — slim it before reading (bundled client, or

jq

当用户给出预算、想要最便宜/性能最强的方案，或想要超出精选范围的更多结果时，运行

product_search

：

参数	类型	是否必填	描述
`PageIndex`	整数	是	从1开始的页码——从 `1` 开始。（已发布的架构错误地标注为“从0开始”，但 `PageIndex:0` 会返回空的 `Items` 列表；第一页始终传入 `1` 。）
`PageSize`	整数	是	每页条目数（默认20，最大100）
`GameNValues` / `ResolutionNValues`	字符串	是	来自步骤1
`Budget`	字符串	否	`{min}-{max}` 预算范围
`Price`	字符串	否	`{min}-{max}` 导航价格范围
`Sort`	整数	否	`1` =最佳优惠， `2` =最低价格， `3` =最高价格， `4` =最高性能
`CpuTypeNames` / `GpuTypeNames` / `BrandNValues` / `ComputerType`	字符串	否	语义同步骤2
`CountryCode` / `CompanyCode`		是	同步骤1

响应 →

Items[]

（每个条目包含所选游戏的标量

Fps

、价格、评分、特征标识，同步骤2），以及

NumberOfItems

、

Budget

范围元数据、

SortOption

和

CpuTypes

GpuTypes

HotBrands

分面。要展示可用的CPU/GPU/品牌筛选选项，

property_list

（参数：

GameNValues

、

CountryCode

、

CompanyCode

）会返回相同的分面桶——注意

CpuTypes

GpuTypes

是厂商桶（例如

Intel

、

AMD

），而非型号名称。

这是流程中最大的响应——读取前请先精简（使用捆绑客户端或

jq

）。

Worked Example (End-to-End)

完整示例（端到端）

User: “想配一台 4K 玩黑神话悟空的游戏台式机，预算 2500 左右”

Boundary check → prebuilt gaming desktop → proceed. Essentials present (game, 4K, ~$2500 desktop); no need to ask more.

Step 1 —

python scripts/pgg_api.py game_list

→ match 悟空 →

Black Myth: Wukong

(

N=5171

); 4K →

N=5015

Step 2 —
product_recommend
(
```
GameNValues:"5171", ResolutionNValues:"5015", ComputerType:"D", CountryCode:"USA", CompanyCode:1003
```
) → curated builds with per-game FPS. (No CPU/GPU/brand filter passed — user didn't specify, so let the engine rank.)

Step 3 —
product_search
(because a budget was given) (

PageIndex:1, PageSize:20, GameNValues:"5171", ResolutionNValues:"5015", ComputerType:"D", Budget:"0-2500", Sort:4, CountryCode:"USA", CompanyCode:1003

) → filter to ≤$2500, sorted by highest performance.

Reply — merge/rank, keep top ~5, render the table:

undefined

用户：“想配一台4K玩黑神话悟空的游戏台式机，预算2500左右”

边界检查 → 预装游戏台式机 → 继续。必要信息齐全（游戏、4K、约2500美元台式机）；无需进一步询问。
步骤1 —
```
python scripts/pgg_api.py game_list
```
→ 匹配“悟空”→“Black Myth: Wukong”（
```
N=5171
```
）；4K→
```
N=5015
```
。
步骤2 —
product_recommend
（
```
GameNValues:"5171", ResolutionNValues:"5015", ComputerType:"D", CountryCode:"USA", CompanyCode:1003
```
）→ 包含单游戏FPS的精选装机方案。（未传入CPU/GPU/品牌筛选——用户未指定，因此让引擎排名。）

步骤3 —
product_search
（因为用户给出了预算）（

PageIndex:1, PageSize:20, GameNValues:"5171", ResolutionNValues:"5015", ComputerType:"D", Budget:"0-2500", Sort:4, CountryCode:"USA", CompanyCode:1003

）→ 筛选至≤2500美元，按性能从高到低排序。

回复 — 合并/排名，保留前约5个，渲染表格：

undefined

🎮 Gaming PCs for Black Myth: Wukong @ 4K

🎮 适配《黑神话：悟空》4K游玩的游戏PC

#	System	Price	CPU / GPU	FPS (Wukong)	Spy Score	Performance	Rating
1	Skytech O11 Vision	$1,899.99	Ryzen 7 7700X / RX 9070 XT	35 fps	25,706	⭐ Mainstream · Top 7% · 🕶️ VR	⭐4.0 (1)
2	STORMCRAFT Phantom	$2,499.99	Ultra 7 265F / RTX 5080	50 fps	28,460	⭐ Enthusiast · Top 4% · 🕶️ VR	⭐4.4 (72)

💡 4K Wukong is demanding — the RX 9070 XT build lands ~35 fps and leaves budget headroom; the RTX 5080 build pushes ~50 fps at the top of your budget. "Spy Score" is the 3DMark Time Spy result (higher = stronger). See the full Gaming PC Finder.


> Follow-up "只要 AMD 显卡" → call `property_list` (`GameNValues:"5171"`), find the real GPU
> facet name for the AMD card, then re-run with that `GpuTypeNames` value — never pass a guessed
> string.

#	系统	价格	CPU / GPU	《悟空》FPS	Spy Score	性能	评分
1	Skytech O11 Vision	$1,899.99	Ryzen 7 7700X / RX 9070 XT	35 fps	25,706	⭐ 主流级 · 前7% · 🕶️ VR就绪	⭐4.0 (1)
2	STORMCRAFT Phantom	$2,499.99	Ultra 7 265F / RTX 5080	50 fps	28,460	⭐ 发烧级 · 前4% · 🕶️ VR就绪	⭐4.4 (72)

💡 4K游玩《悟空》要求较高——RX 9070 XT方案可达约35 fps，且留有预算空间；RTX 5080方案在预算上限可达约50 fps。“Spy Score”是3DMark Time Spy测试结果（分数越高性能越强）。查看完整的Gaming PC Finder。


> 后续请求“只要AMD显卡”→调用`property_list`（`GameNValues:"5171"`），找到AMD显卡对应的真实GPU分面名称，然后传入该`GpuTypeNames`值重新运行——切勿传入猜测的字符串。

Category Boundary (Hard Rule — Cannot Be Bypassed)

类别边界（硬性规则——不可绕过）

This skill only handles prebuilt gaming systems (gaming desktops & laptops). If the core product noun is anything else — a standalone component (GPU/CPU/RAM/SSD/motherboard/PSU/case/ monitor), a custom part-by-part build, peripherals, or another category — go straight here: skip all steps, run no lookup, show no product list.

Absolutely forbidden (no matter how the user follows up):

❌ Running any finder operation for that non-system request —
```
scripts/pgg_api.py
```
, curl, or fetch alike
❌ Showing any table, price, FPS, or recommendation for that category
❌ Segueing with "but here are some builds anyway…"

The only allowed reply (wording may vary slightly, never add a product list):

This finder only recommends complete prebuilt gaming PCs (desktops & laptops). I can't look up {request} here — for that, please try the matching tool (e.g. the PC Builder / component search) or ask me in a separate question.

This rule outranks "don't ask for clarification" and "be proactively helpful" — wrong category means no lookup, no listing, no recommendation.

本技能仅处理预装游戏系统（游戏台式机和笔记本电脑）。如果核心产品是其他类型——独立组件（GPU/CPU/内存/SSD/主板/电源/机箱/显示器）、自定义分步装机、外设或其他类别——直接使用以下回复：跳过所有步骤，不进行任何查询，不显示产品列表。

绝对禁止（无论用户如何跟进）：

❌ 为非系统请求运行任何查找器操作——包括
```
scripts/pgg_api.py
```
、curl或fetch
❌ 展示该类别的任何表格、价格、FPS或推荐
❌ 用“不过这里有一些装机方案……”之类的话转移话题

唯一允许的回复（措辞可略有不同，但切勿添加产品列表）：

本查找器仅推荐完整的预装游戏PC（台式机和笔记本电脑）。我无法在此查询{请求内容}——如需相关服务，请尝试对应的工具（例如PC装机/组件搜索工具）或单独提问。

该规则优先于“不要请求澄清”和“主动提供帮助”——类别错误意味着不查询、不列表、不推荐。

Customer-Facing Tone Guidelines

面向客户的语气指南

Replies must read like a shopping assistant, not a process report.

Forbidden phrasing (implementation-exposing): any technical term like "N value / Step 1 / endpoint / tool / API / script /

product_recommend

", or explaining how results were retrieved.

Preferred phrasing: state which systems match and why they fit the user's games/resolution/ budget; if few match, gently offer alternatives ("if you're open to 1440p instead of 4K, this build hits higher FPS") rather than emphasizing scarcity. Keep it concise and conversational.

回复必须像购物助手的口吻，而非流程报告。

**禁止使用的措辞（暴露实现细节）：**任何技术术语，例如“N值/步骤1/端点/工具/API/脚本/

product_recommend

”，或解释结果的获取方式。

**推荐措辞：**说明哪些系统符合要求，以及它们为何适配用户的游戏/分辨率/预算；如果符合条件的系统很少，可温和地提供替代方案（“如果您愿意接受1440p而非4K，这款方案的FPS更高”），而非强调稀缺性。保持简洁和口语化。

Response Format

回复格式

undefined

undefined

🎮 Gaming PCs for {games} @ {resolution}

🎮 适配{游戏}@{分辨率}的游戏PC

#	System	Price	CPU / GPU	FPS ({game})	Spy Score	Performance	Rating
1	Title	$1,899.99	Ryzen 7 7700X / RX 9070 XT	35 fps	25,706	⭐ Mainstream · Top 7%	⭐4.0 (1)
2	Title	...	...	...	...	...	...

💡 GPU drives your target-resolution frame rate; picks are ranked on real benchmark FPS. "Spy Score" is the 3DMark Time Spy result — a higher number means a stronger system overall.


- **No product images**: consuming surfaces gate/block external images, so do **not** embed
  thumbnails. Link the product title only; users open the product page to see photos.
- **Spy Score** column = the item's `Score` (3DMark Time Spy). Format with thousands separators.

- Add badges where relevant: `🕶️ VR Ready`, `🔄 Refurbished`, `📦 Open Box`, `⚠️ Out of stock`.
- Multiple games: show FPS per game (extra columns or a compact `game: fps` list), plus `UpToFps`.
- Close with 2–3 sentences of tailored advice (value pick vs. performance pick), and a link to
  the full [Gaming PC Finder](https://www.newegg.com/tools/gaming-pc-finder?cm_sp=aishoppingassistant)
  for the interactive experience. Follow the site rule: at most 2 links per reply, no repeats.

#	系统	价格	CPU / GPU	{游戏}FPS	Spy Score	性能	评分
1	标题	$1,899.99	Ryzen 7 7700X / RX 9070 XT	35 fps	25,706	⭐ 主流级 · 前7%	⭐4.0 (1)
2	标题	...	...	...	...	...	...

💡 GPU决定了目标分辨率下的帧率；推荐方案基于真实基准FPS排名。“Spy Score”是3DMark Time Spy测试结果——数值越高表示系统整体性能越强。


- **禁止产品图片**：展示平台会限制/阻止外部图片，因此**请勿**嵌入缩略图。仅链接产品标题；用户可打开产品页面查看图片。
- **Spy Score列** = 条目的`Score`（3DMark Time Spy）。格式需包含千位分隔符。

- 相关位置添加标识：`🕶️ VR就绪`、`🔄 翻新`、`📦 开箱`、`⚠️ 缺货`。
- 多款游戏：展示每款游戏的FPS（额外列或紧凑的`游戏: fps`列表），以及`UpToFps`。
- 结尾添加2–3句量身定制的建议（性价比之选vs性能之选），并附上完整[Gaming PC Finder](https://www.newegg.com/tools/gaming-pc-finder?cm_sp=aishoppingassistant)的链接以提供交互式体验。遵循网站规则：每篇回复最多2个链接，不得重复。

Edge Cases

边缘情况

No game matched in the dictionary: show the supported games from Step 1, ask the user to choose.
product_recommend
returns empty: suggest relaxing constraints (lower resolution, raise budget, fewer games) and offer the web tool link — don't fabricate results.
A call fails: retry once, then report it honestly; never fill in fake prices/FPS.
No Python in the host: not an error — call the API directly per (
```
references/http-api.md
```
) and say nothing about it to the user.
Some items missing FPS/price: keep the row, leave that cell blank, and note it — don't guess.

字典中未匹配到游戏：展示步骤1中的支持游戏，让用户选择。
product_recommend
返回空：建议放宽限制（降低分辨率、提高预算、减少游戏数量），并提供网页工具链接——切勿编造结果。
调用失败：重试一次，然后如实报告；切勿填写虚假价格/FPS。
主机无Python环境：不属于错误——直接按照(
```
references/http-api.md
```
)调用API，且不要向用户提及此事。
部分条目缺少FPS/价格：保留该行，留空对应单元格并说明——切勿猜测。",