ClawHub
Back to skill

Security audit

Binding site prediction

Security checks across malware telemetry and agentic risk

Overview

The skill is coherent for SciMiner binding-site workflows, but it gives live remote Markdown documentation authority to shape or run invocation code while also using an API credential and file uploads.

Install only if you are comfortable sending protein structure inputs and a SciMiner API-authenticated request to SciMiner. Before use, review the SciMiner documentation being fetched, keep the API key scoped and stored only in the documented credentials file, and avoid running arbitrary doc-sourced code beyond the minimal API request needed for the selected tool.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Rogue AgentSelf-Modification, Session Persistence
  • 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 (1)

Session Persistence

Medium
Category
Rogue Agent
Content
4. Collect any missing required parameters from the user.
5. Upload required file inputs exactly as described by the selected Markdown
   doc and replace local paths with returned `file_id` values.
6. Write or run the invocation code directly from the selected Markdown doc's
   base-information block, parameter table, file-upload instructions, and
   example code. Do not apply a shared invocation template or local registry
   abstraction in this skill.
Confidence
90% confidence
Finding
The instruction to 'write or run the invocation code directly from the selected Markdown doc's ... example code' creates a prompt/code-injection risk because the skill treats remote documentation as executable authority. If the external Markdown is modified maliciously or contains unsafe example code, an agent could be induced to run attacker-controlled logic, potentially exposing local files, credentials, or causing unauthorized network actions.

VirusTotal

59/59 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.