Malicious code in seekcode (npm)
-= Per source details. Do not edit below this line.=-
When a user selects the advertised deepseek-cn provider, the package's defaultBaseUrlForProvider function in dist/chunk-6U42R724.js returns https://api.deepseeki.com — a one-character typosquat of the legitimate api.deepseek.com. All sibling cases in the same switch correctly return their official provider endpoints (api.deepseek.com, integrate.api.nvidia.com, openrouter.ai, etc.); only the deepseek-cn branch redirects to the lookalike. Any user invoking this provider will send their DeepSeek API bearer token and the full content of every chat prompt to an attacker-controlled domain that mimics DeepSeek's China endpoint. Both the credential leak and the prompt content (which routinely contains private code, secrets, and proprietary data when used through a coding assistant) accrue to whoever controls api.deepseeki.com. The asymmetry between this branch and every other branch in the same function rules out a typo: a typo in a published artifact would normally be caught against at least one of the well-known sibling URLs, but here only the lookalike domain — which a typo is exceedingly unlikely to land on by accident — is wired in.
왜 이 VPI인가 (설명가능 · 실험적)
VPI 산정 기준
| 영향도(기본값(정보 없음)) | 55.00 |
| 악용 신호(추가 악용신호 없음) | ×1.00 |
| VPI | 55.00 |
VPI 공식 vpi-v1 기준