Security audit

lance-format

Security checks across malware telemetry and agentic risk

Overview

This is a Lance technical reference skill made of markdown documentation, with no executable hooks or hidden runtime behavior, though some copied examples deserve caution.

Installers should treat this as a developer documentation skill. Before reusing examples, replace inline cloud keys with environment variables, IAM or managed identity, or a secret manager, and review any cleanup command such as `rm -rf sift* vec_data.lance` before running it.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (8)

Intent-Code Divergence

Medium

Confidence: 94% confidence
Finding: The document first states that field IDs are immutable, unique, stable, and effectively not reused, but later says dropped column IDs may be reused in some systems. In a format specification, this contradiction can cause implementers to make incompatible assumptions about schema evolution, potentially leading to incorrect field resolution, data corruption, authorization or policy mismatches tied to field IDs, or unsafe cross-version reads.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The documentation shows inline AWS access key, secret key, and session token values directly inside application code without an adjacent warning against hardcoding secrets. Even though these are placeholder values, this pattern normalizes insecure secret handling and can lead users to copy the example into source code, increasing the risk of credential leakage through repos, logs, and build artifacts.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The per-base configuration example includes inline account keys for multiple storage bases and does not warn that embedding credentials in code is insecure. In a storage configuration guide, readers are likely to adapt examples directly, so this can propagate poor secret-management practices across multi-bucket or multi-account deployments.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The Azure example places an account key directly in a Python snippet without cautionary guidance. This is dangerous because it encourages developers to embed long-lived cloud credentials in code, where they may be committed to version control or exposed to anyone with code access.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The AliCloud OSS sample embeds access key and secret key values directly in the example without any nearby security warning. In credential-related documentation, this can cause readers to mirror the pattern, leading to accidental exposure of privileged cloud credentials.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The Volcengine TOS example includes inline access key and secret key fields and lacks warning text about secret exposure. This is a documentation security weakness because it models a dangerous implementation pattern that could result in credential disclosure if copied into real applications.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The Tencent COS example shows secret ID and secret key inline in code, again without warning users not to embed secrets. Because this is setup documentation, users may directly reuse it, increasing the chance of credential compromise through source control, shared notebooks, or support bundles.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The tutorial instructs users to run `rm -rf sift* vec_data.lance` before downloading data, which can delete existing local files matching those patterns without any warning or confirmation. In documentation, destructive shell commands are risky because users often copy-paste them verbatim, and the wildcard broadens the deletion scope beyond only the tutorial artifacts.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.