Back to skill

Security audit

Logseq Bridge

Security checks for vulnerabilities and agentic risk

Overview

This skill is a simple Logseq file-operation guide whose read and write behavior is disclosed and aligned with its purpose, though users should be careful because examples modify local notes.

Install only if you want an agent to read and modify your local Logseq Markdown files. Verify the graph path before running write examples, keep backups or version control for important notes, and be aware that the optional transit-index search may inspect Logseq index files under ~/.logseq rather than only the graph folder you set.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • 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
Findings (6)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill is explicitly designed to read and write a user's local knowledge base, but it presents those capabilities without an upfront warning that commands will modify user data on disk. In a file-writing skill, missing safety language increases the chance of accidental alteration, corruption, or unintended edits to notes, especially when users copy-paste commands directly.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 15)May include surrounding context.

How It Works

text
Agent → Shell → Read/Write local `.md` files → Logseq reloads on next Cmd/Ctrl+R

Pure file operations. No HTTP APIs, no plugin bridges, no port listeners.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The documentation advertises 'pure .md file-based interaction' with 'no plugin bridges' and no extra interfaces, but later includes operations on Logseq's non-Markdown transit database outside the graph directory. This mismatch is security-relevant because it understates the true access scope and may cause users to grant trust under false assumptions about what data will be touched.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill claims to operate on the local Logseq graph via direct Markdown files, but it also instructs searching ~/.logseq/graphs/*.transit, which reaches outside the declared graph root into Logseq's global index/cache area. That expands data access beyond user notes and can expose indexed content from other graphs or metadata the user did not intend to search through this skill.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The quick-start section immediately appends content into the current journal file, which causes a real modification of user notes without any warning, preview, or confirmation step. Even though the content is harmless, this normalizes running write commands directly and can lead to accidental data changes in the wrong graph or file.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The README instructs users to append directly into a Logseq journal file, which modifies user-owned data without any warning, backup guidance, or validation precautions. In a file-manipulation skill, normalizing silent writes to a knowledge base can lead to accidental corruption, unintended content injection, or misuse by downstream agents that follow examples verbatim.

Content

No source excerpt is available for this finding.