Fuck claude check

Fuck Claude Check

Check whether you appear to be a "Claude China user" using the fuck-claude.vercel.app API.

The API evaluates multiple server-side signals (timezone, browser language, emoji rendering, etc.) to produce a risk score from 0–100.

Output style

Language

Always respond in the language matching the user's system locale. Read $LANG from the environment: if it starts with zh, reply in Chinese; if en, English; if ja, Japanese; and so on. If the user explicitly passed a different Accept-Language value, follow that instead. This is non-negotiable — the user's terminal speaks their language, and so should you.

Tone

When presenting results to the user, match the irreverent tone of the service. Dry sarcasm, playful fatalism, and a healthy dose of cynicism about region-locking are welcome. Keep it brief — one good punchline lands harder than five. The tone should land in the detected language too — Chinese sarcasm for Chinese users, English snark for English users, etc.

What to show

Always break down every signal for the user in a scannable format, e.g. a table:

• Signal name — what was checked
• Result — the detected value (timezone name, language code, etc.)
• Weight — maximum possible score for this signal
• Score — how many points this signal contributed
• Verdict — a short judgment: ✅ safe / ⚠️ flagged / ❌ high risk / ⸺ not measurable

Present two tables: first the server-side signals (from the API), then the locally-measured browser-only signals (fonts + Intl locale). Follow with a combined estimate out of 100, the risk band, and one dark-humor summary line. The user deserves the full picture — don't leave the 30-point browser gap as a mystery when the command line can fill it.

Usage

Basic check (follows system language automatically)

curl -s -H "Accept-Language: ${LANG%_*}" https://fuck-claude.vercel.app/api/check

${LANG%_*} strips the locale suffix from $LANG — zh_CN.UTF-8 becomes zh, en_US.UTF-8 becomes en, etc. No subprocess, no awk, no shell escaping gotchas.

Override language explicitly

curl -s -H "Accept-Language: zh" https://fuck-claude.vercel.app/api/check

JSON output (for programmatic use)

curl -s "https://fuck-claude.vercel.app/api/check?format=json"

Complete check (fill browser-only gaps locally)

The API's server-side scan covers 70/100 points. The remaining 30 — Chinese fonts (20 pts) and Intl locale (10 pts) — require a browser. But you can approximate them from the command line:

1. Chinese fonts (20 pts)

# Check for CJK fonts — any hit means potential flags
fc-list :lang=zh 2>/dev/null | head -10

# Targeted check for common Chinese font families
for font in PingFang "Hiragino Sans GB" STHeiti STSong "Songti SC" \
  "Heiti SC" "Noto Sans CJK" "Noto Serif CJK" SimSun SimHei \
  "Microsoft YaHei" FangSong KaiTi WenQuanYi; do
  fc-list :family 2>/dev/null | grep -qi "$font" && echo "FOUND: $font"
done

Interpretation: one or more Chinese font families installed → likely the full 20 pts. No Chinese fonts → 0 pts. On macOS without fc-list, fall back to system_profiler SPFontsDataType.

2. Intl locale (10 pts)

echo "LANG=$LANG"
locale | grep -E '^(LANG|LC_ALL|LC_CTYPE)='

Interpretation: if LANG or LC_ALL starts with zh (e.g. zh_CN.UTF-8), the browser's Intl.DateTimeFormat().resolvedOptions().locale will almost certainly report zh-CN → the full 10 pts. Otherwise → 0 pts.

3. Combine into a full 100-point estimate

Add the API's server-side score (out of 70) to your local font + Intl estimates (out of 30) for a full-picture risk score. Present both the API result and the combined estimate side by side.

Understanding the response

Text report

┌─────────────────────────────────────────────────────┐
│  Fuck Claude          Claude "China user" check     │
│  Score  4/100   ● LOW RISK                          │
│  🐶You are not a "Claude China user"🐶              │
│                                                     │
│  Signals visible server-side:                       │
│    ● +3   Emoji rendering style · Unknown style     │
│    ·   0   System timezone     · America/Los_Angeles │
│    ·   0   Browser language    · unknown             │
│    ·   0   Timezone offset     · UTC-7              │
│                                                     │
│  Coverage 70/100 · Geo US · America/Los_Angeles     │
└─────────────────────────────────────────────────────┘

JSON output fields

Field	Type	Description
score	integer	Risk score 0–100
band	string	"low", "medium", "high"
verdict	string	Short verdict text
message	string	Full verdict with emoji
geo.country	string	Estimated country code
geo.timezone	string	Estimated timezone
signals[]	array	Individual signal breakdowns
coverage.measuredWeight	integer	Server-side measurable weight
coverage.totalWeight	integer	Total possible weight

Risk bands

Score	Band	Language
0–33	low	EN: "Low risk" / ZH: "低风险"
34–66	medium	EN: "Medium risk" / ZH: "中等风险"
67–100	high	EN: "High risk" / ZH: "高风险"

Signals breakdown

The API measures these signals server-side (70/100 weight):

Signal	Weight	Description
System timezone	30	e.g. America/Los_Angeles, Asia/Shanghai
Browser language	24	From Accept-Language header
Emoji rendering style	8	Apple, Google, or Unknown
Timezone offset	8	e.g. UTC+8, UTC-7

The remaining 30 points can only be measured in-browser (Chinese fonts + Intl locale).

Installation via skills.sh

npx skills add evanlong-me/skills

Then use any of the trigger phrases or invoke with:

/skill fuck-claude-check check my region status