ClawHub
Back to skill

Security audit

paper-summary-json

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed academic-paper analysis workflow that copies or downloads user-provided papers and saves extracted text, prompts, JSON, and reports locally.

Install only if you are comfortable with provided papers and derived prompt/report files being stored on the Desktop by default and potentially sent to the model during analysis. Avoid using confidential, unpublished, regulated, or copyrighted-sensitive papers unless you choose an appropriate output directory and understand your model provider’s data handling.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
96% confidence
Finding
The skill instructs the agent to read local files, write outputs, access network URLs, and invoke shell commands, but it declares no permissions or user-facing consent boundary. That mismatch can cause the agent to perform sensitive actions the user may not reasonably expect, especially downloading remote content and writing many artifacts to the Desktop.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The description does not clearly warn that URL inputs will be fetched over the network and that downloaded papers, intermediates, JSON, Markdown, HTML, and DOCX outputs will be persistently stored under the Ubuntu Desktop. This can expose users to privacy, disk usage, and data handling surprises, particularly when papers or URLs are sensitive or when the Desktop is synced or shared.

Ssd 3

Medium
Confidence
92% confidence
Finding
The script writes large portions of extracted paper text, including full body content and evidence spans, into model-facing prompt files on disk. If the analyzed papers contain confidential, copyrighted, unpublished, or regulated data, this design increases the likelihood of unnecessary disclosure to downstream models, logs, operators, or other local users with access to the desktop work directory.

Ssd 3

Medium
Confidence
93% confidence
Finding
The verification prompt again embeds the full original text and asks the model to compare it against extracted results, duplicating sensitive source material in another prompt artifact. This compounds exposure because the same document contents are persisted multiple times and are more likely to be transmitted, logged, or recovered from local storage.

Static analysis

No suspicious patterns detected.