Convert Any Screenshot

• Outcome: Review
• Latest audit: May 27, 2026, 3:44 AM
• Version:
  1.0.1

Overview

The package does not appear malicious, but its advertised screenshot-to-code purpose conflicts with broad generic conversion instructions and unclear write/API-key documentation.

Review carefully before installing. The files do not show exfiltration, destructive actions, or persistence, but the skill should be revised so its instructions, README, credential claims, and modes all match the screenshot-to-code purpose and clearly state whether any local files can be modified.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Description-Behavior Mismatch

Medium

• Confidence: 90% confidence
• Finding: The substantive documentation describes a generic format-conversion tool rather than the declared screenshot-to-code skill. This creates security-relevant ambiguity about expected inputs, processing steps, and outputs, increasing the chance that an agent will route sensitive or unrelated content into the skill under false assumptions.

Intent-Code Divergence

Medium

• Confidence: 89% confidence
• Finding: The in-file intent text directly contradicts the manifest by presenting the skill as a general data-transformation utility. Such contradictory guidance is dangerous because agents often rely on embedded instructions for execution behavior, so this can override user expectations and lead to processing the wrong classes of data.

Intent-Code Divergence

High

• Confidence: 98% confidence
• Finding: The file defines a full test runner but then replaces the effective _main entrypoint with a stub that calls an undefined function, preventing the advertised tests from running. This can silently disable validation and allow malformed or policy-violating skill content to pass review if maintainers assume the auto-generated suite is active.

Missing User Warnings

Medium

• Confidence: 79% confidence
• Finding: Documenting a write mode without warning about file or data modification can lead users to invoke the skill in ways that overwrite local files or alter inputs unexpectedly. In an agent context, unclear write semantics increase the chance of unsafe automation, especially if users assume the operation is read-only or non-destructive.

Vague Triggers

Medium

• Confidence: 78% confidence
• Finding: The activation description is broad enough to match many unrelated transformation requests, which can cause accidental invocation outside the intended screenshot-to-code scenario. Overbroad routing increases exposure to unexpected inputs and compounds the confusion created by the mismatched documentation.

VirusTotal

VirusTotal findings are pending for this skill version.