ClawHub
Back to skill

Security audit

novel-master

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent novel-writing skill that creates local project files for drafts, outlines, and continuity tracking, with no evidence of network exfiltration, credential use, destructive behavior, or hidden execution.

Install only if you are comfortable with the skill storing your manuscript planning data under ~/.qclaw/workspace/novels. Use ordinary book names, review generated files before adding sensitive material, and delete the project directory when you no longer want the draft state retained.

SkillSpector

Vulnerability Patterns
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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)

Lp3

Medium
Category
MCP Least Privilege
Confidence
84% confidence
Finding
The skill instructs creation and ongoing maintenance of files under a local workspace, including state, tracker, and chapter artifacts, but does not declare those file read/write capabilities. Undeclared file access weakens least-privilege guarantees and can lead to unexpected local data modification or overwrite if the skill is auto-invoked or reused in a broader agent environment.

Tp4

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding
The declared description frames the skill as writing assistance, but the body also specifies operational behaviors such as maintaining local workspace state, role-process recovery, graph persistence, and audit gating. This mismatch can mislead users and policy engines about the real effect surface of the skill, increasing the chance of unintended file operations and hidden stateful behavior.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The README advertises helper scripts and a generated project directory under the user's workspace, but it does not clearly warn users that invoking the skill may create and update local files automatically. In an agent setting, undocumented writes and scaffolding can violate user expectations, overwrite existing content, or cause unintended persistence of sensitive manuscript data.

Session Persistence

Medium
Category
Rogue Agent
Content
### Start Writing

Just tell your AI: **"I want to write a mystery novel"**

The AI will automatically:
1. Load genre references → 2. Generate a Creative Brief → 3. Produce Arc Outlines → 4. Write chapters → 5. Track foreshadowing → 6. Self-check quality
Confidence
76% confidence
Finding
The README states that the AI will automatically load references, generate outlines, write chapters, track foreshadowing, and self-check quality, implying persistent multi-step stateful behavior. In combination with tracker files and project state, this can lead to silent session persistence and retained user content without explicit consent or clear lifecycle controls.

Static analysis

No suspicious patterns detected.