ClawHub
Back to skill

Security audit

FEP Alternative

Security checks across malware telemetry and agentic risk

Overview

The skill’s main scientific workflow is understandable, but its bundled helper can make broader web requests and write raw responses more freely than the stated PDB lookup purpose requires.

Install only if you are comfortable with a skill that reads a SciMiner API key, uploads scientific input files to SciMiner, and includes a flexible HTTP helper. Use it for the documented SciMiner and RCSB workflows, avoid passing secrets in helper headers or bodies, and only save raw responses to paths you choose intentionally.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
96% confidence
Finding
The URL builder explicitly permits absolute http/https paths, which turns this script into a general outbound network client rather than a constrained PDB/database retriever. In an agent setting, this can be abused for SSRF-style access to unintended services, data exfiltration to attacker-controlled endpoints, or reaching internal/metadata services if the runtime has network access.

Context-Inappropriate Capability

Medium
Confidence
92% confidence
Finding
The execution path exposes a generic HTTP client with both GET and POST plus attacker-controlled params, headers, and bodies. That capability is substantially broader than the stated scientific retrieval purpose and enables the skill to interact with arbitrary remote services, including sending data out or invoking side-effecting endpoints.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The client forwards arbitrary headers and request bodies without restriction or disclosure, which can carry secrets, bearer tokens, cookies, or sensitive scientific/user data to untrusted destinations. In combination with unrestricted URLs, this materially increases exfiltration and unauthorized action risk.

Missing User Warnings

Medium
Confidence
84% confidence
Finding
Saving raw server responses to an arbitrary path can persist sensitive or unexpected content on disk without safeguards. If the response contains secrets, personal data, or maliciously large content, this can create confidentiality issues, unsafe file placement, or local resource abuse.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.