Security audit

agnes-image-skill

Security checks across malware telemetry and agentic risk

Overview

This skill is a straightforward Agnes image-generation wrapper that sends requested prompts and optional images to Agnes's remote API.

Install this only if you are comfortable sending image prompts and any provided image files or URLs to Agnes's remote service. Keep the AGNES_API_KEY private, confirm the exact image source before image-to-image requests, and verify current pricing and Agnes data-handling terms before using sensitive or proprietary content.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The skill advertises broad trigger phrases like '帮我生成一张图' and similar generic image-related requests, which can easily overlap with ordinary conversation and unrelated image tasks. In an agent environment, this increases the chance of accidental invocation, causing unintended external API calls, cost exposure, and unreviewed transfer of user prompts or image inputs to a third-party service.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The trigger phrase '把这张图改成 XX 风格' is underspecified because many benign editing discussions could match it without the user intending to use this specific skill. In context, the skill can accept local file paths or URLs and send content to an external image API, so a false trigger could expose user-provided images or cause unintended image transformations.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The script sends user-supplied prompts and optional image content to a third-party remote API, but the CLI/help text does not clearly warn users that their text and images leave the local environment. This creates a real privacy and data-handling risk, especially if users provide sensitive images, proprietary content, or personal data under the assumption processing is local.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The trigger phrase "生成一张图" is very generic and likely to match many normal user requests about making an image, which can cause the skill to activate unintentionally. Overly broad activation increases the chance of the wrong tool being invoked, leading to unexpected external API usage, unintended data transfer, or bypass of more appropriate routing logic.

VirusTotal

50/50 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.env_credential_access

Environment variable access combined with network send.

Critical

Code: suspicious.env_credential_access
Location: generate.mjs:37