AI Book Recommendation Engine（AI 书籍推荐引擎）

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:29 PM UTC
• Version:
  1.0.3

Overview

This book recommendation skill is purpose-aligned and disclosed, with a narrow local scoring helper that does not show unsafe behavior.

Before installing, understand that the skill will search the web for your book topics, fetch public book pages, read a narrow reading-history file if present, and run its included scoring script. Review or avoid the helper script if your environment treats any packaged local code execution as unacceptable.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• 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 (2)

Context-Inappropriate Capability

Medium

• Confidence: 84% confidence
• Finding: The skill instructs the agent to execute a local script (`scripts/score_books.py`) even though the stated purpose is a web-search recommendation workflow. Invoking an undeclared local script expands the trust boundary and can lead to execution of unreviewed code, creating a path for unexpected side effects or abuse if that script is modified or malicious.

Scope Creep

High

• Confidence: 94% confidence
• Finding: The instructions reference executing `scripts/score_books.py`, but the manifest only declares read access to `memory/reading-history.json`, so the operational behavior is outside the declared permission model. This mismatch undermines least-privilege guarantees and can hide code execution pathways from reviewers, increasing the risk of unauthorized local actions if the environment permits script execution.

VirusTotal

64/64 vendors flagged this skill as clean.