ClawHub

Bioinformatics Engineer Skill

Security checks across malware telemetry and agentic risk

Overview

This skill is a coherent bioinformatics workflow guide, but users should tighten its example secret-handling and logging templates before using them with real credentials.

Install only if you are comfortable with an agent helping manage bioinformatics projects, local environments, Git/DVC, and project credentials. Before using the templates with real secrets, replace broad `source .env` usage with scoped credential loading and redact output before writing logs, not afterward. Review DVC remote settings and data sensitivity before running `dvc push`.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Intent-Code Divergence

Medium
Confidence
89% confidence
Finding
The skill states that paths should be project-root-relative, but multiple examples and guidance within the same section endorse absolute paths such as /path/to/... and ~/experience-pipelines. This inconsistency weakens the policy and can lead users to embed host-specific or identity-revealing paths into scripts, harming portability and potentially exposing usernames, institutional names, or storage layouts.

Intent-Code Divergence

Medium
Confidence
96% confidence
Finding
The logging template writes all stdout/stderr to the log first via tee, then attempts to scrub the file afterward using a narrow grep pattern that only removes lines starting with a few KEY= names. Any secret echoed in another format, included in command arguments, stack traces, tool debug output, or multiline content will already have been recorded and may remain in the cleaned log, creating a realistic secret-exposure path.

Credential Access

High
Category
Privilege Escalation
Content
cd "$PROJECT_DIR"

# --- 加载密钥 ---
source .env

# --- 激活环境 ---
conda activate {env_name}
Confidence
80% confidence
Finding
.env

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal