Geoskill: Land Use Carbon Accounting

• Outcome: Review
• Latest audit: Jul 31, 2026 · 12:13 PM UTC
• Version: 2.0.0

Overview

The skill mostly performs carbon accounting, but it can download remote imagery and then generate synthetic results that may look tied to the requested location and dates.

Review before installing. Use explicit before/after land-cover rasters for real work, do not treat bbox/date-only output as real carbon accounting, and check whether outbound Planetary Computer queries and download caching are acceptable in your environment. For production or compliance use, require clearer provenance labeling and pinned dependencies.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands

Findings (9)

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 95% confidence
• Finding: The documented purpose is offline carbon accounting over supplied land-cover rasters, but the file also advertises automatic remote data acquisition and synthetic demo generation that materially change trust boundaries and behavior. This is dangerous because users may invoke the skill expecting deterministic local analysis while it may perform network access, fetch external data, or produce outputs from synthetic inputs that could be mistaken for real analytical results.

Intent-Code Divergence

Medium

• Confidence: 92% confidence
• Finding: The Chinese-only section states the skill can automatically download data from Microsoft Planetary Computer, contradicting the earlier CLI and scope that describe supplied local land-cover rasters. Contradictory documentation can mislead operators about what data is used and when network access occurs, undermining informed consent and reproducibility.

Description-Behavior Mismatch

Medium

• Confidence: 93% confidence
• Finding: The documentation expands the skill from local carbon accounting into remote data acquisition, which broadens operational scope and attack surface beyond what users are told in the main description. Hidden or poorly disclosed scope expansion is risky because it can introduce external dependencies, data provenance issues, and privacy concerns without user awareness.

Context-Inappropriate Capability

Low

• Confidence: 84% confidence
• Finding: The skill is described as local carbon accounting, but it depends on a shared data-download library tied to external data sources. This expands the attack surface by introducing network-capable functionality that is not clearly necessary for the stated purpose, increasing supply-chain and unexpected data-access risk if the library is compromised or misused.

Description-Behavior Mismatch

High

• Confidence: 97% confidence
• Finding: The code can present itself as performing carbon accounting from real multi-temporal land-cover inputs while silently falling back to synthetic rasters and explicitly not using the auto-downloaded Sentinel-2 imagery for computation. This creates a serious integrity risk: users may trust fabricated or placeholder results as if they were derived from the requested AOI and dates, leading to false reporting, compliance errors, or policy decisions based on non-real data.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: Automatic data download from an external service is described without a clear warning that invoking the skill may trigger network access and external data transfer. In security-sensitive or regulated environments, undisclosed outbound connections can violate policy, leak location/AOI information, and make executions non-deterministic.

Unpinned Dependencies

Low

• Category: Supply Chain

Content

geoskill-data-fetcher>=0.2.0

# Third-party dependencies used by this skill
numpy
rasterio

• Confidence: 95% confidence
• Finding: numpy

Unpinned Dependencies

Low

• Category: Supply Chain

Content

# Third-party dependencies used by this skill
numpy
rasterio

• Confidence: 94% confidence
• Finding: rasterio

Unpinned Dependencies

Low

• Category: Supply Chain

Content

# Requirements for land-use-carbon-accounting
# Shared data-download library (Planetary Computer / NASA POWER / OSM)
geoskill-data-fetcher>=0.2.0

# Third-party dependencies used by this skill
numpy

• Confidence: 97% confidence
• Finding: geoskill-data-fetcher>=0.2.0

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

No suspicious patterns detected.