cheat-on-content
Audited by Socket on Nov 16, 2026
6 alerts found:
Anomalyx6SUSPICIOUS. 该技能的总体目的与大部分能力基本一致:初始化内容分析项目、建脚手架、导入历史数据、安装本地 hook 都服务于 stated purpose。但其信任边界偏大:会修改 `.claude/settings.json`、落地可执行 shell hook、接触多平台登录态/API key,并自动调用另一个技能;同时关键 adapter 的真实实现与安装来源在本材料中不可验证。未见明确恶意外传、已知窃密端点、隐藏执行或预执行命令,因此不足以判定为恶意,但应按中高风险可疑技能处理。
SUSPICIOUS: the core drafting/brainstorming purpose is coherent, but the footprint is somewhat overbroad and includes medium-risk trust issues. Main concerns are optional installation of a third-party personal skill, forwarding user text/history to that skill, and processing untrusted trend content while retaining Bash/Write/Skill capabilities. No clear malware or credential-harvesting behavior is present.
SUSPICIOUS: 目的与核心能力基本一致,主要是趋势抓取和本地候选池维护;但权限面偏宽(Bash(*)、Skill)、真实抓取命令被下沉到 adapter 文档、且已知相关适配生态包含远程安装/第三方服务依赖。未见明确恶意或直接凭证窃取,但在缺少 adapter 实现细节时应按中等风险处理。
No direct malicious behavior (e.g., credential theft, exfiltration, backdoor logic) is evidenced in this documentation fragment. The principal security concern is the installation pattern that downloads and executes a remote installer script via `curl ... | bash`, which is a common and high-impact supply-chain attack vector if the remote script or hosting is compromised. Verification requires reviewing the actual `install.sh` and the installed skill/router implementation referenced by the documentation.
This configuration sets up asynchronous local OS command execution of a project-local logging script on sensitive user/session-related events (including user prompt submission). While the intent described is benign metadata logging to a local JSONL cache, the actual security posture depends entirely on the unprovided `log-event.sh`. No direct malicious indicators are visible in the fragment itself; however, the combination of (1) shell execution, (2) backgrounding/low observability, and (3) privacy-relevant event triggering results in a moderate supply-chain/security risk that warrants review of the referenced script and any data handling/exfiltration controls it implements.
This fragment configures a synchronous SessionStart hook that executes a project-local shell script via `bash` using `${CLAUDE_PROJECT_DIR}`. While the snippet itself contains no explicit credential theft or network activity, it establishes an arbitrary code execution capability at the start of a session, with behavior entirely dependent on the external `session-start.sh` (and any gating tied to .cheat-state.json). Treat the referenced script and its supply-chain integrity (who can modify it, how it is packaged, and where it is sourced) as the key risk factors.