expert-mode

PassAudited by ClawScan on May 1, 2026.

Overview

This instruction-only skill is coherent and safety-gated, but users should review external expert-list updates before allowing them to persist in memory.

Before installing, understand that confirming updates writes external expert entries into MEMORY.md and can affect future response style. Prefer official sources, inspect previews before confirming, keep auto-sync off unless needed, and avoid sharing sensitive context during update-fetching steps.

Findings (4)

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

Approved expert entries may affect future assistant behavior until the memory entry is changed or removed.

Why it was flagged

The skill persists fetched expert data into agent memory and later uses that data to shape responses. This is expected for the feature, but persistent external context can influence future behavior if a bad update is approved.

Skill content
用户确认后才执行写入 ... 更新MEMORY.md ... 调整自身响应风格以匹配该专家的角色定位
Recommendation

Review update previews carefully, keep expert entries limited to names and descriptions where possible, and remove unexpected MEMORY.md additions.

What this means

An outdated or altered mirror could change the expert list that is proposed for persistence.

Why it was flagged

The update workflow relies on several external repositories and possible mirrors. This is disclosed and purpose-aligned, but mirrored or unofficial sources can differ from the original upstream content.

Skill content
4个仓库 ... GitHub镜像 - fastgit、ghproxy等国内镜像 ... Gitee镜像 - 搜索Gitee同步镜像
Recommendation

Prefer the official GitHub repositories, inspect the proposed changes before confirming, and avoid accepting updates from unknown mirrors.

What this means

If used, additional automation or sub-agents may participate in fetching update content.

Why it was flagged

The skill describes fallback fetching through browser automation and spawned sub-agents. This is tied to retrieving public expert-list updates, but it introduces delegated tool or agent activity users should notice.

Skill content
浏览器自动化 - 浏览器可继承系统代理 ... 子代理抓取 - spawn子代理尝试不同网络
Recommendation

Use these fallback methods only when needed and avoid including sensitive task context in update-fetching prompts.

NoteHigh Confidence
ASI10: Rogue Agents
What this means

If enabled, the skill may periodically check for updates and ask to write them to memory.

Why it was flagged

The skill includes a scheduled auto-sync option. It is disabled by default and still requires confirmation before writing, so this is disclosed and bounded persistence rather than hidden background behavior.

Skill content
Cron自动同步功能默认禁用 ... 需要用户主动开启 ... 开启后每次同步仍需确认
Recommendation

Leave auto-sync disabled unless you need it, and verify each proposed update before confirming.