ClawHub
Back to skill

Security audit

academic-paper-writer

Security checks for vulnerabilities and agentic risk

Overview

This academic writing skill is coherent and purpose-aligned, but users should be aware it can query Zotero/PubMed and may offer to add references to Zotero.

Install this only in the intended JD workspace. Before using it with confidential or unpublished research, confirm whether Zotero/PubMed queries are acceptable, use least-privilege Zotero credentials, and require explicit confirmation before adding any DOI to a Zotero collection.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
Findings (4)

Description-Behavior Mismatch

Medium
Confidence
90% confidence
Finding
The skill’s declared purpose is drafting/revision and citation lookup, but it also instructs the agent to add new items into Zotero collections. That expands the capability from read-oriented assistance to persistent modification of the user’s reference library, which can cause unintended state changes, metadata pollution, or insertion of low-quality references if triggered without explicit confirmation.

Context-Inappropriate Capability

Low
Confidence
74% confidence
Finding
The skill includes direct calls to the Zotero API using an API key, which introduces secret-bearing external access and possible disclosure of user research topics in outbound queries. Although related to citation lookup, the use of raw API access is broader and riskier than necessary if not clearly scoped, sanitized, and disclosed.

Vague Triggers

Medium
Confidence
83% confidence
Finding
The trigger phrases are broad enough that the skill may activate during ordinary writing-related conversation, causing unintended retrieval actions or workflow changes. In context, accidental activation could lead to unsolicited external searches or citation operations based on incomplete or sensitive draft content.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The skill instructs queries to Zotero and PubMed without clearly warning that user-supplied research topics, hypotheses, or draft details may be transmitted to external services. In an academic context, that can leak unpublished ideas, sensitive project direction, or other confidential research information beyond the local environment.

Static analysis

No suspicious patterns detected.