Security audit

PSD Batch Export

Security checks across malware telemetry and agentic risk

Overview

This is a coherent local PSD batch-export skill, with some privacy and scoping cautions but no artifact-backed malicious behavior.

Install only if you are comfortable letting the skill read PSDs, spreadsheets, fonts, and generated images and write batch outputs/reports. Avoid optional LLM mode for confidential customer or event data unless you approve sending prompts to your configured provider, and treat OCR verification reports as sensitive local files.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Intent-Code Divergence

Medium

Confidence: 89% confidence
Finding: The advertised layer-based replacement is implemented by mapping layer names to current text and then patching by matching original text content globally. If multiple layers share identical text, the wrong layer may be modified, enabling unintended edits to generated design assets and undermining template integrity in batch workflows where precise field placement matters.

Vague Triggers

Medium

Confidence: 78% confidence
Finding: The description contains very broad activation phrases such as 'Use whenever the user mentions PSD' and other generic design-related terms, which can cause the skill to trigger in contexts broader than necessary. Overbroad auto-selection increases the chance that a high-capability skill with Read/Write/Edit/Bash access is invoked for loosely related requests, expanding the attack surface and enabling unintended file or command operations.

Missing User Warnings

Low

Confidence: 91% confidence
Finding: The README directs users to place downloaded font files into a directory that the skill automatically scans and uses, but it does not clearly warn users that anything added there becomes active input to the workflow. In this PSD automation context, that increases the chance users will ingest unvetted third-party font files from external sources, which expands the attack surface and can lead to supply-chain or malformed-font parsing risks in downstream tooling.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: User-provided descriptions, layout elements, and template context are sent to external OpenAI/Anthropic-compatible APIs without an explicit notice, consent flow, or data-classification check. In a PSD workflow, prompts may include customer names, event details, internal branding, or other sensitive design content, so silent transmission creates a real privacy and data-governance risk.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: When OCR mode is enabled, the script extracts expected text from PSD text layers and recognized text from rendered PNG regions, then writes those values into the JSON report. In a PSD batch-export workflow, those layers may contain names, IDs, tickets, certificates, invitations, or other sensitive production data, so the report can become an unintended secondary store of sensitive content without any masking, consent prompt, or retention control.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.