Ralph Ultra Security Audit

• Outcome: Pass
• Latest audit: May 27, 2026, 10:03 PM
• Version:
  3.0.0

Overview

This is a broad but instruction-only security audit skill; it can inspect sensitive project details and write a local report, but those actions fit its stated purpose.

Install only if you want a long, invasive security review of a repository or system you are authorized to audit. Set scope before running, require approval before any active probing, and keep .ralph-report.md private because it may contain vulnerabilities, environment details, or secret locations.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• 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
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Vague Triggers

Medium

• Confidence: 92% confidence
• Finding: The skill description includes broad trigger phrases such as 'before major release' and 'security incident investigation' that can match ordinary project conversations and cause unintended invocation. Because this skill is designed to perform extensive repository and environment inspection over many iterations, accidental activation increases the chance of unnecessary sensitive data exposure and disruptive actions.

Missing User Warnings

Medium

• Confidence: 88% confidence
• Finding: The instructions direct the agent to rename existing reports and auto-save to `.ralph-report.md`, but the skill description does not warn users that execution will modify files in the workspace. Hidden write behavior is risky in security-audit contexts because it can overwrite artifacts, interfere with existing reports, or create persistence of potentially sensitive findings without informed consent.

Missing User Warnings

Low

• Confidence: 84% confidence
• Finding: The auto-detect steps instruct the agent to inspect git metadata, stack files, infrastructure manifests, and CI/CD configuration, but the skill description does not clearly warn that repository structure and environment details will be examined. In a security-audit skill this behavior is contextually expected, but the lack of upfront disclosure still creates a transparency and consent problem, especially where configuration files may contain sensitive operational details.

VirusTotal

66/66 vendors flagged this skill as clean.