Back to skill

Security audit

obsidian

Security checks for vulnerabilities and agentic risk

Overview

The verdict is benign with medium confidence: the scanners found Obsidian vault actions and screenshot capture that are not fully reflected in the skill description, and the scan was partial.

The verdict is benign, but consider whether you want the skill to move or delete notes and save an application screenshot. The scanner found no executable scripts and AIG reported no findings, but Skillspector marked its analysis partial because some local path references were unresolved.

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
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

The code uses the declared CLI and operates on the expected vault resources, but its commands perform relocation and deletion rather than the listed operations. In particular, deleting a note is a distinct, undeclared capability rather than a supporting implementation detail of reading, searching, creating, or editing notes.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
86% confidence
Finding

Plugin reload and counting vault files are consistent with the declared scope, and developer error inspection can support plugin management. However, capturing the application UI introduces an undeclared data-capture capability beyond the described note, task, link, property, and plugin operations. The supplied JavaScript only counts files; it does not demonstrate broader data access or exfiltration.

Content

No source excerpt is available for this finding.