Security audit

DeepRead OCR

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only DeepRead OCR API integration, but users should treat uploaded documents, webhooks, BYOK providers, and preview links as sensitive data-sharing paths.

Install only if you are comfortable sending selected documents to DeepRead and, when enabled, to your webhook endpoint or BYOK model provider. Do not process confidential documents unless that external sharing is allowed by your policy, keep DEEPREAD_API_KEY out of files and chats, secure webhook receivers, and treat preview URLs as private links because anyone with the link may be able to view document previews or extracted data.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (5)

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The skill encourages users to upload potentially sensitive documents to a third-party OCR service but does not clearly warn that both source files and extracted contents leave the local environment. In a document-processing context, this omission is security-relevant because invoices, contracts, receipts, and forms frequently contain PII, financial, or confidential business data.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The webhook examples instruct users to send OCR results to an arbitrary external callback URL without warning that processed document contents and extracted fields will be pushed to that endpoint. This can lead to unintentional exfiltration if users configure third-party, misconfigured, or non-secure webhook receivers.

Missing User Warnings

High

Confidence: 99% confidence
Finding: The documentation presents preview URLs and an unauthenticated public preview endpoint without a prominent warning that anyone possessing the link may access document contents and extracted data. For OCR workflows involving contracts, invoices, and forms, this creates a meaningful confidentiality risk through link leakage, logs, browser history, chat sharing, or referrer exposure.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The BYOK section says processing routes through the user's provider account but does not clearly warn that document data may also be transmitted to OpenAI, Google, OpenRouter, or other connected model providers. Users may incorrectly assume only DeepRead handles the data, leading to unreviewed third-party disclosure and compliance issues.

Ssd 3

High

Confidence: 99% confidence
Finding: This is more than a missing warning: the skill explicitly promotes unauthenticated public preview links for processed documents. Because the preview may expose original document images and extracted structured data, any leaked or guessed link can disclose sensitive records without user authentication.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.