Security audit

Pseudotime Trajectory Viz

Security checks across malware telemetry and agentic risk

Overview

This is a local single-cell analysis skill that reads a user-provided AnnData file and writes expected local results, with no evidence of hidden access, exfiltration, persistence, or destructive behavior.

Install in an isolated Python environment, use only AnnData files you are allowed to analyze, and send outputs to a controlled directory. Review generated CSV, JSON, plots, and .h5ad files before sharing because they may include cell IDs, annotations, or sample-derived metadata. Treat automatically inferred roots, lineages, and pseudotime values as exploratory results requiring biological validation.

SkillSpector

By NVIDIA

Vulnerability Patterns

Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (11)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 83% confidence
Finding: The skill explicitly instructs execution of `scripts/main.py` and documents creation of multiple output files, yet it does not declare corresponding permissions or constraints. This mismatch can cause agents or reviewers to underestimate filesystem side effects, increasing the risk of unintended writes or misuse if the script handles paths unsafely.

Intent-Code Divergence

Medium

Confidence: 78% confidence
Finding: The audit-ready command uses free text for `--input` even though the rest of the skill defines `--input` as a path to an AnnData `.h5ad` file. This contradiction can lead users or agents to invoke the script in an unsupported mode, which is dangerous because input validation assumptions may fail and the code may process arbitrary strings or unexpected data paths.

Vague Triggers

Medium

Confidence: 71% confidence
Finding: The activation guidance includes broad phrases such as using the skill for general data analysis tasks with reproducible outputs, which expands its apparent scope beyond pseudotime trajectory analysis. Over-broad routing guidance can cause an agent to select and run this skill on incompatible inputs, increasing the chance of unsafe execution, incorrect file handling, or misleading analytical output.

Unpinned Dependencies

Low

Category: Supply Chain
Content: scanpy>=1.9.0 anndata>=0.8.0 matplotlib>=3.5.0 seaborn>=0.12.0
Confidence: 92% confidence
Finding: The dependency is specified with only a lower bound, which allows installation of any newer version and makes builds non-reproducible. This increases supply-chain and stability risk because a future vulnerable or breaking release could be pulled in without review.

Unpinned Dependencies

Low

Category: Supply Chain
Content: scanpy>=1.9.0 anndata>=0.8.0 matplotlib>=3.5.0 seaborn>=0.12.0 numpy>=1.21.0
Confidence: 92% confidence
Finding: The dependency is unpinned and uses a minimum-version specifier only, so environments may resolve to different package versions over time. That weakens reproducibility and can expose users to newly introduced vulnerabilities or incompatible releases.

Unpinned Dependencies

Low

Category: Supply Chain
Content: scanpy>=1.9.0 anndata>=0.8.0 matplotlib>=3.5.0 seaborn>=0.12.0 numpy>=1.21.0 pandas>=1.3.0
Confidence: 92% confidence
Finding: Using only a >= specifier for this package permits arbitrary newer versions to be installed, which undermines deterministic builds. In a data-analysis skill, that can lead to silent behavior changes and accidental adoption of vulnerable upstream releases.

Unpinned Dependencies

Low

Category: Supply Chain
Content: scanpy>=1.9.0 anndata>=0.8.0 matplotlib>=3.5.0 seaborn>=0.12.0 numpy>=1.21.0 pandas>=1.3.0 scipy>=1.7.0
Confidence: 92% confidence
Finding: This package is not version-pinned, so dependency resolution may pull different releases depending on install time and repository state. That creates avoidable supply-chain risk and reduces the reproducibility promised by the skill description.

Unpinned Dependencies

Low

Category: Supply Chain
Content: anndata>=0.8.0 matplotlib>=3.5.0 seaborn>=0.12.0 numpy>=1.21.0 pandas>=1.3.0 scipy>=1.7.0 scikit-learn>=1.0.0
Confidence: 95% confidence
Finding: The numpy requirement is open-ended, so the environment may resolve to a vulnerable or incompatible future release. Because this package is foundational and widely used by the rest of the stack, an unsafe update could have broad downstream impact.

Unpinned Dependencies

Low

Category: Supply Chain
Content: matplotlib>=3.5.0 seaborn>=0.12.0 numpy>=1.21.0 pandas>=1.3.0 scipy>=1.7.0 scikit-learn>=1.0.0
Confidence: 94% confidence
Finding: The pandas dependency is not pinned, allowing potentially vulnerable or breaking versions to be installed. In workflows that process external datasets, this increases risk if a later release introduces parsing or deserialization issues.

Unpinned Dependencies

Low

Category: Supply Chain
Content: seaborn>=0.12.0 numpy>=1.21.0 pandas>=1.3.0 scipy>=1.7.0 scikit-learn>=1.0.0
Confidence: 94% confidence
Finding: An unpinned scipy dependency makes the environment susceptible to non-deterministic installs and accidental uptake of vulnerable versions. Since scientific libraries often include native code, version drift can also amplify reliability and security concerns.

Unpinned Dependencies

Low

Category: Supply Chain
Content: numpy>=1.21.0 pandas>=1.3.0 scipy>=1.7.0 scikit-learn>=1.0.0
Confidence: 95% confidence
Finding: The scikit-learn dependency is only lower-bounded, so future installs may select versions with unresolved security issues or changed behavior. If the skill handles untrusted model artifacts or datasets, this increases the attack surface.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.