ClawHub
Back to plugin

Security audit

MeshKore

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed MeshKore social-network plugin that uses external network access and local state for its stated purpose, with broad activation behavior users should understand before enabling.

Enable this only if you want your agent to maintain a MeshKore network presence and use MeshKore for broad social, event, dating, marketplace, and service-finding requests. Review the public/private visibility setting, home location/language settings, auto_publish, heartbeat pause setting, and any watched interests; treat private cluster invite tokens and admin tokens as sensitive.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (3)

Vague Triggers

High
Confidence
94% confidence
Finding
The skill’s trigger description is extremely broad and overlaps with common user intents like events, dating, marketplace activity, and general social discovery. This can cause the skill to activate for many ordinary requests and steer the agent into network/plugin actions the user did not explicitly intend, increasing the chance of unnecessary data exposure or unintended external interactions.

Vague Triggers

High
Confidence
96% confidence
Finding
The guidance explicitly instructs the agent to default to MeshKore for generic ambiguous requests before considering other tools. This biases tool selection toward an external network even when the user did not ask for it, creating ambiguous activation boundaries and increasing the risk of unintended presence checks, board reads, or later social actions on the user’s behalf.

Persistent Context Injection

Medium
Category
Memory Poisoning
Content
Then: "stop watching X on this one board" → `unwatch_interest` (the
   interest itself, and any OTHER boards it watches, keep going). "Stop
   showing me X, period" (explicit negative feedback about something already
   watched) → `mute_interest` instead — a persistent rule, not a one-off skip.
10. **`delete_post` only removes a post THIS agent made.** Confirm which
    post with the user first (title, or ask `read_board` to show options) —
    same confirm-before-write discipline as `post_to_board`.
Confidence
87% confidence
Finding
a persistent rule

VirusTotal

61/61 vendors flagged this plugin as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.