Security audit

Thumbnail QA

Security checks across malware telemetry and agentic risk

Overview

The skill appears purpose-built for thumbnail QA, but it can automatically edit a repository and create git commits from broad or automatic triggers.

Install only if you are comfortable with an agent inspecting your Next.js project, starting a local dev server, editing CSS, writing .gstack reports/screenshots, and making git commits. Use it on a clean branch and require a preview or explicit confirmation before applying or committing fixes.

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: 92% confidence
Finding: The skill’s stated purpose is thumbnail QA, but it also changes repository hygiene and runtime state by appending to `.gitignore`, creating an unrelated commit, and potentially launching background processes. This expands the skill’s side effects beyond user-visible image fixes, creating a mismatch between declared scope and actual behavior that can lead to unintended repository modifications.

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The skill promises atomic commits for thumbnail fixes, but also performs a separate standalone commit for `.gitignore` setup, which is unrelated to any individual image correction. This can pollute project history, violate user expectations, and normalize hidden or incidental commits under the guise of a narrowly scoped automation task.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The changelog explicitly advertises that the skill will automatically fix code and create atomic commits, but the skill description shown to users does not prominently warn that running it can modify the repository and mutate git history. This creates a consent and safety gap: a user may invoke what sounds like a QA/checking tool without realizing it performs write actions, increasing the chance of unintended code changes or commits in sensitive repositories.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The trigger phrases are broad everyday-language commands like "Check my thumbnails" and "Fix image cropping," and the README also says the skill runs automatically after image uploads. In an agent environment, this increases the chance of accidental invocation during ordinary conversation or routine asset changes, which can unexpectedly launch browsing, analysis, code edits, and commits without a tightly scoped request.

Missing User Warnings

Medium

Confidence: 98% confidence
Finding: The README advertises automatic fixes with before/after evidence and states that each fix gets its own atomic commit, but it does not describe any explicit confirmation or approval step before modifying code or creating commits. In this skill's context, that is especially risky because the workflow can browse pages, compute changes, edit source, and persist those changes to version control, amplifying the impact of accidental invocation or incorrect analysis.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.