Skill flagged β€” suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

TokenSaver Korean

v2.0.0

🧠 ν•œκ΅­μ–΄ Context DB for AI Agents β€” 토큰 91% 절감. μž„λ² λ”© 의미 검색, 쀑볡 μžλ™ 병합, λ©”λͺ¨λ¦¬ 계측화(Hot/Warm/Cold), μžλ™ μ•„μΉ΄μ΄λΈŒ, WAL ν”„λ‘œν† μ½œ, μ—”ν‹°ν‹° μΆ”μΆœ. Korean-first + English support.

⭐ 0· 94·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for dorongss/token-saver-korean.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "TokenSaver Korean" (dorongss/token-saver-korean) from ClawHub.
Skill page: https://clawhub.ai/dorongss/token-saver-korean
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Required env vars: FIREWORKS_API_KEY
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Canonical install target

openclaw skills install dorongss/token-saver-korean

ClawHub CLI

Package manager switcher

npx clawhub@latest install token-saver-korean
Security Scan
Capability signals
Requires sensitive credentials
These labels describe what authority the skill may exercise. They are separate from suspicious or malicious moderation verdicts.
VirusTotalVirusTotal
Suspicious
View report β†’
OpenClawOpenClaw
Suspicious
medium confidence
β„Ή
Purpose & Capability
The declared purpose (Korean-first context DB with embedding search, dedup, tiers, WAL) aligns with the code and requirements: the client uses an embedding API and NLP libs (konlpy/nltk). Requesting FIREWORKS_API_KEY matches the EmbeddingManager which posts to api.fireworks.ai. However some README/installation hints (npm steps for an otherwise Python package) and the presence of a platform-specific initializer script suggest sloppy packaging or environment assumptions.
!
Instruction Scope
SKILL.md describes generic initialization and 'automatic workspace detection', but init_bora_context.py contains hard-coded Windows paths (C:\Users\Roken\.openclaw\workspace\...) and will read local files (MEMORY.md, USER.md, SOUL.md, AGENTS.md, IDENTITY.md) and then call client.save_memory() and generate embeddings. That means local workspace documents could be read and (via the embedding API) transmitted externally; this behavior is not explicitly surfaced in SKILL.md and is scope-creep from a user-perspective.
βœ“
Install Mechanism
There is no remote download or opaque installer: dependencies are standard Python packages (konlpy, nltk) referenced in SKILL.md and pyproject. No suspicious external URL downloads or extract operations were found. The only network call in code is to the Fireworks embeddings endpoint which is expected for an embedding-based skill.
β„Ή
Credentials
Only FIREWORKS_API_KEY is required, which is proportionate to using an external embedding provider. However the key would be used to send content (possibly sensitive) to the fireworks.ai API; requiring a single embedding API key is reasonable but the combination with the initializer that reads local files increases the risk of unintended data transmission. No other unrelated secrets are requested.
βœ“
Persistence & Privilege
The skill does not request always:true and does not attempt to modify other skills or system-wide configs. It stores embeddings and some JSON files locally (embeddings.json, wal.json) which is expected for a local context DB. The initializer modifies/reads local workspace files only (no evidence of escalating privileges), but that local access is environment-specific and should be reviewed.
What to consider before installing
This skill appears to implement what it claims, but exercise caution before installing or running it: - Inspect and (if needed) remove or edit init_bora_context.py before running. It contains hard-coded Windows paths that read several workspace files and then save them into the memory DB; these files could contain sensitive info and will be sent (for embedding) to the external Fireworks API if a key is present. - Only provide a FIREWORKS_API_KEY you trust. Treat that key as granting the embedding service access to any text you send. - If you want to try the skill safely, run it in an isolated environment or test VM, and do not export your real workspace paths or secrets into that environment. - Confirm you are happy with local storage of embeddings.json/wal.json and where client.save_memory persists data. - The README contains an odd npm-related instruction for installing a Python packageβ€”review packaging/install steps carefully before copying files into ~/.openclaw or running automated install commands. If you want higher assurance, ask the author for a version of the initializer that prompts before reading files or accepts explicit file paths, and a minimal README that documents exactly which local paths the skill will read and what data it will send to api.fireworks.ai.

Like a lobster shell, security has layers β€” review code before you run it.

Runtime requirements

EnvFIREWORKS_API_KEY
latestvk977agtmvqjavd6pvapm3x4a7h854430
94downloads
0stars
1versions
Updated 1w ago
v2.0.0
MIT-0

🧠 TokenSaver Korean v2.0

AI μ—μ΄μ „νŠΈλ₯Ό μœ„ν•œ ν•œκ΅­μ–΄ μ΅œμ ν™” Context Database

토큰 μ†Œλͺ¨λ₯Ό μ΅œλŒ€ 91% μ ˆκ°ν•˜λŠ” 슀마트 λ©”λͺ¨λ¦¬ μ‹œμŠ€ν…œ. μ €μž₯Β·κ²€μƒ‰Β·μ••μΆ•Β·κ³„μΈ΅ν™”Β·μ•„μΉ΄μ΄λΈŒκΉŒμ§€ μ™„μ „ μžλ™ν™”.


✨ 핡심 κΈ°λŠ₯

κΈ°λŠ₯ν•œ 쀄 μ„€λͺ…
πŸ” μž„λ² λ”© 의미 κ²€μƒ‰ν‚€μ›Œλ“œκ°€ μ•„λ‹ˆλΌ 의미둜 μ°ΎμŠ΅λ‹ˆλ‹€
πŸ”— 쀑볡 μžλ™ λ³‘ν•©λΉ„μŠ·ν•œ λ©”λͺ¨λ¦¬ μ•Œμ•„μ„œ ν•©μ³μ€λ‹ˆλ‹€
πŸ“Š λ©”λͺ¨λ¦¬ 계측화Hot β†’ Warm β†’ Cold μžλ™ 관리
πŸ—„οΈ μžλ™ μ•„μΉ΄μ΄λΈŒ30일 묡은 λ©”λͺ¨λ¦¬ μžλ™ 정리
πŸ“ WAL ν”„λ‘œν† μ½œμ €μž₯ λ¨Όμ €, 응닡 λ‚˜μ€‘ β€” μœ μ‹€ 제둜
🏷️ μ—”ν‹°ν‹° μΆ”μΆœμ‚¬λžŒΒ·λΈŒλžœλ“œΒ·μ œν’ˆ μžλ™ νƒœκΉ…
🌐 λ‹€κ΅­μ–΄ μ§€μ›ν•œκ΅­μ–΄(konlpy) + μ˜μ–΄(nltk)

πŸš€ λΉ λ₯Έ μ‹œμž‘

1. ν™˜κ²½λ³€μˆ˜ μ„€μ •

# Fireworks API ν‚€ (μž„λ² λ”© κ²€μƒ‰μš©)
export FIREWORKS_API_KEY="your-key"

# Windows PowerShell
$env:FIREWORKS_API_KEY="your-key"

2. Pythonμ—μ„œ μ‚¬μš©

from token_saver.client import TokenSaverKorean

# μ΄ˆκΈ°ν™” (μžλ™μœΌλ‘œ workspace 탐지)
client = TokenSaverKorean()

# μ €μž₯ β€” μž„λ² λ”© μžλ™ 생성 + 쀑볡 μ‹œ μžλ™ 병합
client.save_memory(
    uri="biz/daily_sales",
    content="였늘 맀좜 500λ§Œμ›, ROAS 5.83",
    category="biz"
)

# 검색 β€” ν‚€μ›Œλ“œ + μž„λ² λ”© ν•˜μ΄λΈŒλ¦¬λ“œ
results = client.find("맀좜 ν˜„ν™©", use_embedding=True)

# ν‹°μ–΄ 톡계
stats = client.get_tier_stats()
# β†’ {'hot': 14, 'warm': 0, 'cold': 0, 'archive': 0}

# 만료 λ©”λͺ¨λ¦¬ 정리
client.cleanup_expired(days=30)

3. OpenClawμ—μ„œ μ‚¬μš©

λ§ˆμŠ€ν„°: "λ‹₯ν„°λ ˆμ΄λ”” ROAS κΈ°μ–΅ν•΄"
β†’ 보라: memory_store(uri="memories/drlady/roas", content="ROAS 5.83...")

λ§ˆμŠ€ν„°: "μ €λ²ˆμ— λ§ˆμΌ€νŒ… μ „λž΅ λ­μ˜€μ§€?"
β†’ 보라: memory_recall(query="λ§ˆμΌ€νŒ… μ „λž΅")

πŸ“Š 토큰 절감 효과

μ‹œλ‚˜λ¦¬μ˜€κΈ°μ‘΄ 토큰TokenSaver절감율
전체 Context λ‘œλ“œ50,0004,50091%
ν•œκ΅­μ–΄ 검색20,0002,00090%
λ©”λͺ¨λ¦¬ μ••μΆ•30,0003,50088%

πŸ—‚οΈ λ©”λͺ¨λ¦¬ 계측화

μžλ™μœΌλ‘œ μ€‘μš”λ„μ— 따라 ν‹°μ–΄λ₯Ό κ΄€λ¦¬ν•©λ‹ˆλ‹€:

ν‹°μ–΄κΈ°κ°„μ €μž₯ 방식
πŸ”₯ Hot7일 내전체 λ‚΄μš© + μž„λ² λ”©
🌑️ Warm30일 λ‚΄μš”μ•½ + μž„λ² λ”©
❄️ Cold30일+ν‚€μ›Œλ“œλ§Œ (μ••μΆ•)
πŸ—„οΈ Archive30일 λ―Έμ ‘μ†κ²€μƒ‰μ—μ„œ μ œμ™Έ
# μžλ™ 승격/κ°•λ“± β€” μ ‘κ·Όν• μˆ˜λ‘ Hot으둜 올라감
client.save_memory("test", "λ‚΄μš©")  # β†’ Hot
# 30일 ν›„ μžλ™ β†’ Cold
# λ‹€μ‹œ κ²€μƒ‰ν•˜λ©΄ β†’ Warm β†’ Hot

πŸ” ν•˜μ΄λΈŒλ¦¬λ“œ 검색

ν‚€μ›Œλ“œ λ§€μΉ­ + μž„λ² λ”© μœ μ‚¬λ„λ₯Ό κ²°ν•©ν•œ ν•˜μ΄λΈŒλ¦¬λ“œ 검색:

results = client.find(
    query="κ΄‘κ³  효율 κ°œμ„ ",
    limit=5,
    use_embedding=True    # μž„λ² λ”© 검색 ON
)

# κ²°κ³Ό: ν‚€μ›Œλ“œ 점수 + μž„λ² λ”© 점수 = μ΅œμ’… μˆœμœ„
# "ROAS μ΅œμ ν™” 방법" (12.45점)
# "κ΄‘κ³  μ†Œμž¬ μ„±κ³Ό 뢄석" (9.91점)

API μ‹€νŒ¨ μ‹œ μžλ™μœΌλ‘œ ν‚€μ›Œλ“œ κ²€μƒ‰μœΌλ‘œ 폴백 β€” μ ˆλŒ€ μ•ˆ λŠκΉ€.


πŸ”— 쀑볡 μžλ™ 병합

λΉ„μŠ·ν•œ λ‚΄μš©μ΄λ©΄ μƒˆλ‘œ λ§Œλ“€μ§€ μ•Šκ³  κΈ°μ‘΄ 것에 μ—…λ°μ΄νŠΈ:

# 첫 μ €μž₯
client.save_memory("biz/roas", "ROAS 5.83", category="biz")

# λΉ„μŠ·ν•œ λ‚΄μš© μž¬μ €μž₯ β†’ μžλ™ 병합 (코사인 > 0.85)
client.save_memory("biz/roas", "ROAS 6.12둜 μƒμŠΉ", category="biz")
# β†’ 같은 URI에 버전 μ—…κ·Έλ ˆμ΄λ“œ (v1 β†’ v2)

🏷️ μžλ™ μ—”ν‹°ν‹° μΆ”μΆœ

μ €μž₯ μ‹œ μ‚¬λžŒΒ·λΈŒλžœλ“œΒ·μ œν’ˆμ„ μžλ™μœΌλ‘œ 인식:

entities = client.extract_entities("κΉ€λͺ…μ§„ λŒ€ν‘œκ°€ λ‹₯ν„°λ ˆμ΄λ”” 리μ₯¬-ν†‘μŠ€ μΆœμ‹œ")
# β†’ {'persons': ['κΉ€λͺ…μ§„'], 'brands': ['λ‹₯ν„°λ ˆμ΄λ””'], 'products': ['리μ₯¬-ν†‘μŠ€']}

βš™οΈ μ„€μ • 파일 (ovk.conf)

{
  "embedding": {
    "dense": {
      "provider": "fireworks",
      "model": "qwen3-embedding-8b",
      "dimension": 768
    }
  },
  "language": "auto",
  "token_optimization": {
    "enabled": true,
    "target_reduction": 0.91
  }
}

πŸ›‘οΈ μ•ˆμ •μ„±

κΈ°λŠ₯μ„€λͺ…
WAL ν”„λ‘œν† μ½œμ‘λ‹΅ 전에 λ¨Όμ € μ €μž₯ β€” ν¬λž˜μ‹œν•΄λ„ μœ μ‹€ μ—†μŒ
μžλ™ λ°±μ—…μˆ˜μ • μ „ 원본 λ°±μ—… 보관
μž„λ² λ”© 폴백API μž₯μ•  μ‹œ ν‚€μ›Œλ“œ κ²€μƒ‰μœΌλ‘œ μžλ™ μ „ν™˜
ν•˜μœ„ν˜Έν™˜v1 데이터 κ·ΈλŒ€λ‘œ μ‚¬μš© κ°€λŠ₯

πŸ“¦ μ˜μ‘΄μ„±

νŒ¨ν‚€μ§€μš©λ„ν•„μˆ˜?
konlpyν•œκ΅­μ–΄ ν˜•νƒœμ†Œ λΆ„μ„κΆŒμž₯
nltkμ˜μ–΄ NLPꢌμž₯
FIREWORKS_API_KEYμž„λ² λ”© κ²€μƒ‰μž„λ² λ”© μ‚¬μš© μ‹œ

konlpy/nltk 없이도 μž‘λ™ν•©λ‹ˆλ‹€ (κΈ°λ³Έ ν† ν¬λ‚˜μ΄μ € μ‚¬μš©).


πŸ“„ λΌμ΄μ„ μŠ€

MIT License β€” 자유둭게 μ‚¬μš©, μˆ˜μ •, 배포 κ°€λŠ₯.


TokenSaver Korean v2.0 μž„λ² λ”© 검색 Β· 쀑볡 병합 Β· λ©”λͺ¨λ¦¬ 계측화 Β· μžλ™ μ•„μΉ΄μ΄λΈŒ ν•œκ΅­μ–΄ μš°μ„ , μ˜μ–΄ 지원, OpenClaw μ΅œμ ν™”

Comments

Loading comments...