Mongolian AI for Codex

• Outcome: Pass
• Latest audit: Jul 28, 2026 · 3:18 PM UTC
• Version: 2.0.1

Overview

This skill clearly works by sending user-selected Mongolian language inputs to a stated external API, with appropriate warnings for sensitive content and no hidden persistence found.

Install only if you are comfortable sending selected text, images, audio, and documents to mongol.open-idea.net under your API key. Do not use it for confidential, regulated, credential-containing, or highly personal material unless you explicitly approve that upload and understand the billing implications.

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (8)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 88% confidence
• Finding: The skill declares capabilities to read environment variables, read files, and execute shell commands, but the finding indicates those permissions are not explicitly declared. That creates a transparency and governance gap: users or platforms may not realize the skill can access local files and secrets or invoke external commands, increasing the risk of unintended data exposure or abuse if the skill is compromised or misused.

Natural-Language Policy Violations

Medium

• Confidence: 88% confidence
• Finding: The file instructs the skill to use `mw` as the default language and `mn` for Cyrillic without requiring explicit user confirmation. Defaulting locale can cause unintended processing in the wrong language variant, reduce user control, and potentially send user content to an external service under assumptions the user did not approve.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The document translation instructions tell the agent to upload local Word/PDF files to a remote API and only note file size, format, and response handling, without reiterating in this file that the document contents leave the local environment. Because Word/PDF files often contain sensitive business or personal data, this can cause unintentional exfiltration if users or downstream agents follow the instructions without explicit consent and data-sensitivity review.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: This file documents sending chat content, images, audio, and documents to an external third-party API using bearer-authenticated requests, but it does not itself require a user-facing disclosure or confirmation step at the point of transmission. Because these payloads may contain sensitive personal or confidential data, the omission can lead to unintended data exfiltration or privacy violations if the skill sends user content without explicit informed consent.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The script sends user text or full message history to an external API endpoint, but this script itself provides no runtime warning or confirmation before transmitting potentially sensitive content. In this skill’s context, that matters because conversations may include private text, and the metadata explicitly warns not to send sensitive or confidential data without explicit confirmation.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: This script uploads the full user-supplied document to a remote service via curl using multipart form data, but the script itself provides no runtime warning, confirmation, or sensitivity check before transmission. In the context of a translation skill handling DOCX/PDF files, this creates a real data-exfiltration/privacy risk because users may pass confidential documents without realizing the contents leave the local environment.

Missing User Warnings

Medium

• Confidence: 97% confidence
• Finding: The script base64-encodes the entire local image and sends it to a remote API endpoint, which is a real data-exfiltration/privacy risk if users provide sensitive screenshots, scans, or identity documents. Although the skill metadata mentions not sending sensitive data without confirmation, this script itself provides no user-facing warning, consent prompt, or sensitivity check before transmitting the file contents.

External Transmission

Medium

• Category: Data Exfiltration

Content

*) die "unsupported image extension: $EXTENSION" ;;
esac

require_commands curl python3
load_key

HEADER_FILE=$(mktemp)

• Confidence: 95% confidence
• Finding: This script performs external transmission by preparing a JSON request containing the full base64-encoded image and sending it with curl to a remote service. In the context of an OCR skill, such transmission is expected functionality, but it remains security-relevant because local file contents may include confidential information and are exported off-host.

VirusTotal

65/65 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.