Security audit

ADME Property Predictor

Security checks for vulnerabilities and agentic risk

Overview

This skill is a local small-molecule ADME prediction helper with some documentation and dependency hygiene issues, but no evidence of hidden, destructive, credential-seeking, or unrelated behavior.

Reasonable to install if you need local, rough ADME estimates for small molecules. Use only valid SMILES strings or CSV files with SMILES data, run it in a virtual environment, pin or review dependencies if reproducibility matters, and do not treat its outputs as experimental or regulatory-grade evidence.

SkillSpector

Vulnerability Patterns

Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
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

Findings (4)

Intent-Code Divergence

Medium

Confidence: 89% confidence
Finding: The documentation includes example invocations using clinical-style free text despite the skill being a molecule/SMILES-based ADME predictor. That contradiction can mislead an agent into feeding arbitrary text into the tool, producing nonsensical outputs, fallback behavior, or accidental downstream misuse of fabricated-looking results as if they were valid pharmacokinetic predictions.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The 'When to Use' section is broad enough to attract generic data-analysis tasks outside the tool's real scope. In an agent setting, overbroad routing guidance is dangerous because it increases the chance that the skill is selected for inappropriate tasks, where it may fail open, generate irrelevant outputs, or trigger unintended script execution on unsuitable inputs.

Unpinned Dependencies

Low

Category: Supply Chain
Content: dataclasses rdkit
Confidence: 97% confidence
Finding: The dependency 'dataclasses' is unpinned, so builds may resolve to different versions over time and introduce unexpected or malicious changes through the supply chain. In this specific case the risk is somewhat reduced because 'dataclasses' is a well-known package and, on modern Python versions, often unnecessary as it is part of the standard library, but leaving it unpinned still harms reproducibility and dependency integrity.

Unpinned Dependencies

Low

Category: Supply Chain
Content: dataclasses rdkit
Confidence: 99% confidence
Finding: The dependency 'rdkit' is unpinned, which allows future installs to pull different package versions and can introduce breaking changes, compromised releases, or transitive dependency risk. Because this skill performs chemistry/ADME analysis and likely relies heavily on RDKit behavior, unpinned resolution also undermines reproducibility and could materially affect analytical outputs in addition to supply-chain exposure.

Static analysis

No suspicious patterns detected.