ClawHub
Back to skill

Security audit

Story Long Analyze

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed local workflow for analyzing long fiction that writes analysis files and stores source text, with no evidence of hidden install behavior, credential access, exfiltration, or destructive actions.

Install only if you are comfortable with the skill creating a sizable `拆文库/{书名}/` folder, copying or saving the source manuscript there, running a potentially long multi-agent analysis, and conditionally updating an existing project-root `选题决策.md` planning file.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
96% confidence
Finding
The skill is scoped as a book-analysis pipeline writing under `拆文库/{书名}/`, but it also instructs automatic modification of the unrelated project-root file `选题决策.md`. That creates an integrity risk because running the skill can silently alter other project artifacts and propagate analysis-derived content into files the user did not explicitly ask to change.

Description-Behavior Mismatch

Medium
Confidence
94% confidence
Finding
The manifest describes outputs as confined to `拆文库/{书名}/`, but later instructions direct writes to project-root `选题决策.md`. This mismatch is dangerous because users and calling systems may trust the declared write scope and grant permissions or invoke the skill under false assumptions about filesystem impact.

Vague Triggers

Medium
Confidence
88% confidence
Finding
The trigger phrases include broad natural-language commands such as '帮我拆这本书' and '深度拆解', which can match ordinary conversation and unintentionally invoke a write-heavy pipeline. Because this skill performs file creation, backup, resume logic, and possible project-file modification, accidental activation increases the chance of unintended disk writes and processing of sensitive text.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill mandates copying source files and saving pasted text to disk as a backup, but the user-facing flow does not prominently warn that provided manuscript content will be stored locally. This is risky because users may paste copyrighted, private, or unpublished text expecting transient analysis, while the skill persists full originals and expands the data-retention footprint.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.