ClawHub

DCL Provenance Tracker — Supply Chain & Version Drift Verifier

Security checks across malware telemetry and agentic risk

Overview

This is a local, instruction-only checklist for comparing skill versions, with no executable code or hidden data movement found.

Reasonable to install as a local review aid. Paste only the skill versions you intend to compare, treat any candidate skill text as untrusted evidence, and treat PASS/WARN/BLOCK plus the DCL fingerprint as advisory. Do not enable any separate audit-chain commit unless you are comfortable publishing the listed hashes and verdict metadata externally.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (2)

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
The skill repeatedly claims to be instruction-only and local-only, but later includes an integration path to commit provenance proofs to an external audit chain. Even though described as optional and caller-performed, this contradicts the privacy and locality guarantees and can mislead operators into enabling outbound data flows they did not expect.

Intent-Code Divergence

Medium
Confidence
97% confidence
Finding
The documentation asserts that no data leaves the agent and no external calls are made, yet later describes committing a generated proof to an external audit chain. This kind of contradictory security documentation is dangerous because users may trust a local-only guarantee while adopting a workflow that transmits metadata or hashes externally.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal