Security audit

Doc Guard Free

Security checks for vulnerabilities and agentic risk

Overview

The skill is an encrypted notes tool, but it needs review because it broadly routes database tasks and encourages returning full decryptable share links.

Install only if you trust the MCP server you configure. Treat full share links as secrets because they include decryption material, avoid using this skill for SQL or general database tasks, and require explicit confirmation before deletion or any export/import/reset-style operation.

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

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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Vague Triggers

High

Category: Not specified by scanner
Confidence: 97% confidence
Finding: The skill advertises itself for generic database operations, SQL queries, and storage management even though the body describes an encrypted Markdown document tool. This overly broad trigger can cause an agent to invoke the skill in unrelated data contexts, increasing the chance of inappropriate tool use, unintended data access attempts, or mishandling sensitive database tasks under false assumptions about end-to-end encryption.
Content: No source excerpt is available for this finding.

Missing User Warnings

High

Category: Not specified by scanner
Confidence: 97% confidence
Finding: The guidance to always return full share links is dangerous because the link contains the #encryptionKey fragment, which the document itself identifies as key material required for decryption. Encouraging unconditional return or propagation of that full link increases the risk of exposing decryptable access to sensitive documents through chat logs, tool traces, screenshots, history, or unintended recipients, undermining the skill's privacy model.
Content: No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium

Category: Not specified by scanner
Confidence: 92% confidence
Finding: The manifest description says the core scope is limited to create, read, update, list, and basic search. However, the skill file later documents deletion as a core ability and also mentions export, import, and reset style operations, which materially expand the behavior beyond the stated five scenarios for the free edition.
Content: No source excerpt is available for this finding.

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 92% confidence
Finding: The capability scope is defined as a broad bag of keywords, including generic and mismatched terms, instead of precise activation rules. In an agent ecosystem this can lead to over-selection of the skill for loosely related prompts, causing sensitive user requests to be routed to a tool that may not be appropriate for the task and increasing the risk of unintended data handling.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 84% confidence
Finding: The examples show a direct deletion command ('删除文档 abc123') even though deletion is irreversible. While later sections mention confirmation, the example itself omits any warning and could normalize executing destructive actions without first informing the user.
Content: No source excerpt is available for this finding.

Intent-Code Divergence

Low

Category: Not specified by scanner
Confidence: 77% confidence
Finding: The file states the free version supports only single-document processing per operation, yet elsewhere it documents listing recent documents, paginated listing, and searching documents as normal usage. Those instructions contradict the stated limitation because they inherently operate across multiple documents.
Content: No source excerpt is available for this finding.