Security audit

Protein Key Fragment Analysis

Security checks across malware telemetry and agentic risk

Overview

This appears to be a local protein-analysis skill with some documentation and scientific-claim caveats, but no hidden data access, persistence, or malicious behavior.

Install only if you need local protein consensus/key-fragment analysis and have clustalo available. Treat generated functional and disulfide-bond claims as predictions requiring biological validation, and be aware that parts of the documentation mention optional combined aa-pair workflows that are not included in this package.

SkillSpector

By NVIDIA

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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 (5)

Description-Behavior Mismatch

Medium

Confidence: 78% confidence
Finding: The documented output structure includes unrelated amino-acid-pair analysis artifacts, formulation files, and cross-tool combined reporting, which exceeds the stated purpose of protein key-fragment analysis. This scope drift is dangerous because it can cause the skill to invoke or rely on additional workflows, data sources, and outputs that users and policy systems did not expect.

Description-Behavior Mismatch

Medium

Confidence: 80% confidence
Finding: The 'species comprehensive report' explicitly includes amino-acid pair frequency analysis and Top-5 pair details, which are outside the declared scope of protein key-fragment and consensus analysis. This mismatch increases the risk of over-broad invocation and hidden coupling to other tools, leading to unreviewed processing steps and unexpected data handling.

Intent-Code Divergence

Medium

Confidence: 93% confidence
Finding: The report states that the conserved cysteines 'form disulfide bonds' as a factual conclusion, while the method section only claims detection of potential disulfide-forming cysteines. This is an overclaim that can mislead downstream users into treating an unvalidated structural inference as established, which is especially problematic in a scientific analysis pipeline meant to generate structured reports.

Intent-Code Divergence

Medium

Confidence: 91% confidence
Finding: The report labels two conserved cysteines as a disulfide-bond feature, but the cited evidence only shows conserved Cys positions and does not establish bond formation. In a bioinformatics analysis skill, this is dangerous because it can mislead downstream functional interpretation, structural claims, or experimental design by presenting speculation as validated evidence.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The CLI matchers are broad and generic enough to trigger on common protein-analysis requests beyond this skill’s narrow workflow, which can cause inappropriate auto-selection and execution in contexts the user did not intend. In an agent ecosystem, overbroad routing is dangerous because it can mis-handle user data, produce misleading scientific results, or bypass a more appropriate specialized skill.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.