Security audit

Reference Formatter

Security checks for vulnerabilities and agentic risk

Overview

This reference-formatting skill is purpose-aligned and contains no executable payload, but users should be aware it may create output files and optionally use web lookup for bibliography details.

Install only if you are comfortable with the agent creating a references document in the workspace. For private bibliographies, internal URLs, or annotated references, ask the agent to keep processing local and to get your confirmation before any web lookup or file write.

A.I.G

Vulnerability Patterns

Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
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 (5)

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The trigger list contains broad, common phrases such as '引用格式' and 'citation format' that may match ordinary conversation and invoke the skill unintentionally. In a skill that can write files and optionally query external services, over-triggering increases the chance of unintended processing, file creation, or privacy-impacting actions.

Intent-Code Divergence

Medium

Confidence: 97% confidence
Finding: The workflow says to directly generate the document and attach the change report without user confirmation, which contradicts the changelog claim that confirmation is required before document generation. This can cause the agent to take side-effecting action earlier than the user expects, creating files or processing content without an explicit approval step.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The skill defaults to writing output files to the workspace root and names them automatically, but the main workflow does not clearly warn the user before creating files. Silent or weakly disclosed file creation is risky because it introduces side effects, may overwrite user expectations about where data goes, and can expose sensitive bibliography contents in shared workspaces.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The optional web lookup for incomplete references can transmit bibliography entries, including potentially sensitive internal titles, author lists, URLs, or annotations, to external services without any privacy warning. Even though lookup is described as optional, the absence of a disclosure/consent step creates a real risk of unintended data exfiltration.

Natural-Language Policy Violations

Low

Confidence: 71% confidence
Finding: The description text is primarily in Chinese but also embeds English usage instructions (“Use when: ...”), implying a mixed-language default presentation. The file does not state that the skill should match the user’s preferred language or offer an opt-in choice.

Static analysis

No suspicious patterns detected.