ClawHub
Back to skill

Security audit

Scent Trails

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-only coordination protocol with strong privacy limits; the optional blockchain discussion needs careful implementation but is not active behavior.

Safe to install as a conceptual/documentation skill. If you implement storage, shared fields, witness records, or blockchain anchoring from these ideas, keep them opt-in, consent-based, non-identifying, and avoid storing raw user content, identifiers, logs, or reconstructable history.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The section introduces optional on-chain anchoring and mentions irreversible transformation/privacy-preserving patterns, but it does not clearly warn users that blockchain data, metadata, and linkability can remain publicly observable and difficult to revoke in practice. In a skill or design document, this can mislead implementers into believing that meaning can decay while privacy risk is fully mitigated, creating real exposure through permanent inference, correlation, or attribution.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.