Promptify Skill

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only prompt optimizer whose codebase search, web research, and clipboard behavior are disclosed and aligned with its purpose.

Install only if you are comfortable with a prompt optimizer that may inspect relevant project files for +deep or auto-detected codebase prompts, may send prompt-derived terms to web search for +web or current-practice prompts, and may produce clipboard commands. Review prompts containing secrets or private project details before using those modes.

SkillSpector

By NVIDIA

Vulnerability Patterns

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

Findings (8)

Vague Triggers

Medium

Confidence: 85% confidence
Finding: The auto-trigger examples are broad enough that normal conversational phrases like 'Help me with this thing' or 'Add auth to our API' could unintentionally activate clarifying questions, codebase research, or web search behaviors. In a prompt-optimization skill that may route to sub-agents and external capabilities, unintended invocation can expose context, trigger unnecessary data access, or cause surprising actions the user did not explicitly request.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The README states that the skill 'Copies to clipboard' but provides no warning, consent flow, or disclosure about when clipboard writes occur. Clipboard contents are security-sensitive because silent overwrites can replace user data, facilitate accidental pasting of modified prompts elsewhere, or create confusion in workflows involving secrets or production commands.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The description uses broad, common phrases such as 'improve this prompt' and 'make this clearer' as invocation cues, which can cause the skill to trigger in situations where the user did not intend to invoke it. In an agentic environment, over-broad routing increases the chance of unintended prompt transformation, context exposure to sub-agents, or execution of auxiliary behaviors like research dispatch.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The auto-detection signals rely on vague everyday language like 'best practices', 'make it better', or references to 'this project', which are highly context-dependent and prone to false positives. Because these signals can trigger clarifier, codebase, or web-research sub-agents, ambiguous matching can expand access to additional tools or context without sufficiently clear user intent.

Missing User Warnings

Low

Confidence: 94% confidence
Finding: The skill directs copying generated content to the system clipboard via `pbcopy` without warning or obtaining user consent. Clipboard writes can overwrite sensitive user data currently stored there and may place confidential prompt content into a shared OS resource that other applications or workflows can access.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The trigger phrases are broad enough to match many ordinary development requests such as integrating, extending, or refactoring existing systems. In a prompt-optimization skill that can route into codebase exploration, this can cause unintended invocation and unnecessary access to repository context, increasing the chance of over-collection of sensitive internal code or instructions.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The trigger phrases are broad enough to match common user language such as 'best practices' or 'latest', which can cause the web-researcher agent to activate when web access is unnecessary. This creates unnecessary external lookups, increases data exposure to web/search tools, and may route tasks away from safer or more appropriate local-only handling.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The description and metadata advertise a broad, generic capability to optimize prompts and include multiple modifiers that expand behavior, which can cause the skill to trigger in loosely related contexts. Over-broad invocation increases the chance the agent routes sensitive or unrelated user input into this skill, potentially causing unintended codebase reads or web access when combined with modifiers.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal