Bohrium Dataset Management

• Outcome: Pass
• Latest audit: May 28, 2026 · 9:27 AM UTC
• Version: 1.0.0

Overview

This skill is a disclosed Bohrium dataset-management helper, but it should be used carefully because it handles access keys and can change or delete remote datasets.

Install only if you intend to let the agent help manage Bohrium datasets. Store ACCESS_KEY in a secure environment or secret manager, do not commit real keys into config files, verify dataset IDs before mutations, require explicit confirmation before deletes or version changes, and review the remote bohr CLI installer source before running curl-to-bash commands.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 85% confidence
• Finding: The skill clearly uses sensitive capabilities: network access to Bohrium endpoints and environment-based credential access, yet it declares no permissions or capability requirements. That mismatch can cause operators or orchestrators to grant implicit access without visibility, increasing the chance of unintended credential use or outbound requests.

Description-Behavior Mismatch

Medium

• Confidence: 86% confidence
• Finding: The code exposes dataset permission inspection even though the manifest description does not mention that capability. This mismatch can cause users or orchestrators to invoke a more privacy-sensitive function than expected, increasing the chance of unauthorized metadata disclosure or policy bypass through misleading scope.

Missing User Warnings

Medium

• Confidence: 88% confidence
• Finding: The authentication section shows direct placement of an access key in configuration and environment variables but gives no guidance on secret handling, redaction, storage, or rotation. This raises the risk of accidental credential exposure through logs, screenshots, checked-in config files, or shell history, which could allow unauthorized API access.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The skill documents irreversible deletion commands without recommending confirmation, scoping checks, backups, or dry-run verification. In an agent setting, destructive commands presented this plainly can lead to accidental data loss if invoked on the wrong dataset IDs or without clear user confirmation.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: Creating a new dataset version is a remote state-changing action, but the code performs it immediately with no explicit confirmation, preview, or safety interlock. In an agent context, this increases the risk of accidental or unauthorized modifications triggered by ambiguous prompts or chaining mistakes.

Env Variable Harvesting

High

• Category: Data Exfiltration

Content

```python
import os, requests

AK = os.environ.get("ACCESS_KEY", "")
BASE = "https://open.bohrium.com/openapi/v1/ds"
HEADERS = {"accessKey": AK}
HEADERS_JSON = {**HEADERS, "Content-Type": "application/json"}

• Confidence: 82% confidence
• Finding: os.environ.get("ACCESS_KEY

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

No suspicious patterns detected.