Security audit

logseq

Security checks across malware telemetry and agentic risk

Overview

This skill is a legitimate Logseq helper, but its examples expose broad note-editing power without enough safeguards.

Install only if you are comfortable giving an agent broad access to your local Logseq graph. Use it read-only by default, back up the graph first, avoid the generic HTTP bridge unless it is bound to localhost with authentication and an allowlist, and require explicit confirmation before writes, deletes, moves, Git commands, or bulk automation.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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 (4)

Context-Inappropriate Capability

Medium

Confidence: 89% confidence
Finding: The documented bridge plugin exposes a generic method dispatcher over HTTP that forwards arbitrary method names and arguments into the Logseq API. In practice, this can turn a local note-taking interface into an unauthenticated command surface capable of reading, modifying, or deleting graph content if the bridge is reachable by other local processes or a browser context.

Missing User Warnings

Medium

Confidence: 80% confidence
Finding: The skill documents content creation, update, and deletion capabilities against the user's Logseq graph without warning that these operations are destructive and may permanently alter local notes. In an agent-skill context, omission of safety guidance increases the chance that an automation uses these APIs on behalf of the user without clear confirmation or safeguards.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: These examples demonstrate creating pages, inserting blocks, updating content, and moving blocks without clearly warning that they modify the user's local knowledge graph. In this skill context, the API operates against a live local Logseq instance, so copying examples directly can cause unintended data loss, duplication, or corruption if users assume they are read-only snippets.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The automation examples perform bulk actions such as archiving completed tasks and auto-tagging pages across the graph, but they lack user-facing cautions or safeguards. In a Logseq skill tied to a local live graph, wide-scope mutations are more dangerous because a user may run them verbatim and unintentionally reorganize or alter large portions of personal notes.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.