quant-buddy-skill

该技能整体更像真实的金融数据/量化分析技能，能力与用途大体一致，主数据流也指向声明的官方服务，因此不属于明显恶意。但它要求本地读取 API Key、允许本地脚本执行、写入 session/输出文件，并包含对外公式包发布能力，权限与数据流范围较简单“查股价”场景更宽，故应评为可疑偏中风险而非恶意。

This code is primarily a tool runner with session management and output post-processing. It contains no clear, explicit malware behaviors in this fragment (no eval/exec payloads, no obvious exfil to hardcoded suspicious domains, no reverse shells). The main security concern is supply-chain/integrity risk from the auto self-update mechanism executed via subprocess based on remote version-check results, plus an unvalidated @file path that can lead to arbitrary file read/overwrite. Overall, it appears more like a legitimate orchestrator with notable security gaps and an update vector that warrants deeper review of config.json endpoint trust and especially self_update.py.

No definitive evidence of intentional malware/backdoor behavior is present in this fragment (no exec/eval, no subprocess-based command execution, no stealthy credential harvesting beyond using an intended API key). However, there are meaningful security risks: (1) uploadData can transmit arbitrary local file contents provided via file_path to the configured remote endpoint (high-impact exfiltration risk), and (2) saveChart writes to an output path derived directly from untrusted params['name'] without sanitization (path traversal/unintended overwrite risk). These issues should be reviewed/mitigated, especially under scenarios where CLI parameters or endpoint configuration are not fully trusted.