Turn Any Book into a Working Agent 一键把书变成员工

• Outcome: Review
• Latest audit: May 29, 2026 at 9:23 PM UTC
• Version:
  1.0.0

Overview

The skill has a clear book-to-agent purpose, but it can create persistent skill files from book-derived names without enough containment or activation controls.

Run this only in a clean folder containing the intended book. Review the generated SKILL.md before using or publishing it, and rename generated agents to a simple safe name without slashes, dot-dot paths, shell metacharacters, or instruction-like text.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Vague Triggers

Medium

• Confidence: 84% confidence
• Finding: The activation phrase "请作为《{book_title}》专家帮助我..." is broad and likely to overlap with ordinary user conversation, which can cause accidental invocation of the generated skill. In agent systems that route or prioritize skills based on trigger phrases, ambiguous activation increases the risk of unintended behavior, misrouting, or prompt/scope confusion.

Vague Triggers

Medium

• Confidence: 86% confidence
• Finding: The generated SKILL instructions tell users to activate the agent with a vague natural-language phrase rather than a constrained command. This makes collisions with normal requests more likely and can unintentionally activate the generated role in contexts where the user did not intend to delegate to that skill.

Vague Triggers

Medium

• Confidence: 82% confidence
• Finding: The example activation phrase uses a generic request pattern that resembles ordinary conversation, making accidental triggering plausible. In ecosystems with many skills, this can lead to skill collision and unintended role adoption, which is a real security and safety concern when capabilities differ across skills.

Vague Triggers

Medium

• Confidence: 80% confidence
• Finding: The Windows example repeats the same vague activation pattern, reinforcing an unsafe invocation model across platforms. Consistently broad triggers increase the chance that downstream generated skills will be invoked unintentionally, especially when copied verbatim by users or developers.

VirusTotal

64/64 vendors flagged this skill as clean.