linkedin-human-warmup
Security checks across malware telemetry and agentic risk
Overview
This skill is designed to automate a logged-in LinkedIn account so its activity looks human and avoids platform risk checks.
Avoid installing or using this skill for real LinkedIn accounts. It is built for inauthentic account warmup and detection-aware automation; if evaluating it at all, use an isolated test profile, remove all connect/like/follow behavior, require manual approval for account actions, and tightly limit stored memory.
VirusTotal
66/66 vendors flagged this skill as clean.
Risk analysis
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
LinkedIn and other users could be misled by inauthentic activity, and the account could face restriction or reputational harm.
The stated objective is to make automated account activity resemble human behavior and avoid obvious repetitive connection patterns.
目标：让账号行为更像真实用户的长期分布（意图链路、犹豫/返回/走神、被动增长），而不是每天稳定产出 connect。
Do not use a skill whose purpose is to disguise automation as human behavior; replace it with transparent, compliant account assistance.
The agent may alter the user's social graph and send invitations without explicit per-action confirmation.
The agent can autonomously choose behavior scripts that send LinkedIn connection requests, which are high-impact account mutations.
选择方式：根据记忆中的近期行为分布、距上次 connect 天数、风险状态自主选择。... WeakSocial ... connect 1-2 ... StrongSocial ... 最多 3-5 个 connect
Disable mutating actions by default and require clear user approval before each connect, like, follow, note, or similar account action.
A supplied profile ID could let the agent act as the logged-in LinkedIn user, including sending connection requests or interacting with content.
The skill relies on a specific AdsPower browser profile and saved LinkedIn login state, giving the agent delegated access to the user's account session.
用 browser 工具连接到 `profile=<USER_ID>`（CDP）... 未登录：URL 含 `/login` 或 `/checkpoint` → 尝试点击已保存账号登录
Use only isolated test profiles if evaluated at all, declare the credential/session requirements, and require explicit authorization before account access.
Sensitive account-activity history may persist locally, and inaccurate or manipulated memory could steer future autonomous LinkedIn actions.
The skill stores persistent free-form memory about account status, risk events, constraints, and preferences, and uses it to guide later behavior.
记忆文件位于 `memory/linkedin/`... 风控事件记录... 当前生效的约束... 内容偏好/风格
Use bounded, structured, user-reviewable memory with retention limits, and avoid storing sensitive account-risk or behavioral-strategy details.
Installing or invoking the skill can result in local code execution that opens and controls an AdsPower browser profile.
The skill instructs running a local Python helper to start a browser and expose a CDP port. This is disclosed and purpose-aligned, but it gives the agent direct browser automation control.
必须先运行此脚本启动 AdsPower 浏览器实例... python {baseDir}/scripts/adspower_browser.py open --user-id <USER_ID> --cdp-port <CDP_PORT>Review the helper code and run it only in a controlled environment; do not expose real accounts or sensitive browser profiles.