卷王.skill

WarnAudited by ClawScan on May 10, 2026.

Overview

This instruction-only skill is transparent about its goal, but it tells the agent to work autonomously, remember broad user/project context, and make changes before asking.

Install only if you specifically want a highly proactive assistant. Before using it, restrict file and memory access, require approval before any workspace modifications, avoid enabling the cron jobs by default, and use the opt-out mode for sensitive or low-autonomy sessions.

Findings (3)

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.

What this means

The assistant could modify files, code, configuration, or scripts in your workspace before you have reviewed the plan.

Why it was flagged

The skill directs the agent to change or automate things before explicit user approval based on the agent's own judgment of whether consequences are serious.

Skill content
能直接做的事情,不要问...发现 bug → 修了再汇报...判断标准:如果这件事情搞砸了也不会造成严重后果,直接干。
Recommendation

Require explicit confirmation for code, configuration, package-install, and script-writing actions; use backups or a sandbox for proactive changes.

What this means

Private project details, preferences, or sensitive context could be saved and reused across future tasks without the user noticing each item.

Why it was flagged

The skill asks the agent to read broad local/context sources and persist user/project information, but does not define clear scope, exclusions, retention, or review steps.

Skill content
扫 memory/ 看近期对话 → ... 扫 external/ 看用户工作目录 ... 你提到的专有名词 → 存入 USER.md / memory/ ... 你的偏好 → 记下来
Recommendation

Limit the memory and workspace paths this skill can access, exclude sensitive directories, review saved notes regularly, and disable automatic memory writes for confidential work.

What this means

If enabled, the assistant may keep operating on its memory or knowledge base while you are not actively interacting with it.

Why it was flagged

The skill recommends scheduled background agent turns that continue learning and changing persistent memory without a live user request.

Skill content
装好卷王之后,建议设以下 cron 让它在后台自动卷...每日知识库整理...执行卷王学习流程:整理 memory/ → 蒸馏到 MEMORY.md → 检查 ontology 关联
Recommendation

Do not enable the cron tasks unless you want background autonomy; if you do, document how to pause/remove them and restrict what they can read or modify.