agent-security-audit

• Outcome: Review
• Latest audit: Jul 25, 2026 · 8:32 AM UTC
• Version: 1.2.4

Overview

This security-audit skill is review-worthy because its documentation promises local/static scanning, but the bundled script is a SaaS client that sends supplied code to a remote HTTP API.

Install only if you are comfortable with code being sent to the publisher's remote audit service. Do not submit proprietary code, secrets, credentials, regulated data, or private repositories unless the endpoint, transport security, retention policy, and vendor trust are acceptable to you; prefer local-only tools for sensitive reviews.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• 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

Findings (5)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 89% confidence
• Finding: The skill declares required environment variables and documents outbound API usage, but it does not declare corresponding permissions/capabilities in the skill interface. This can mislead reviewers and users about the skill’s effective access to secrets and networked data transfer, reducing transparency around data exposure risk.

Vague Triggers

Medium

• Confidence: 83% confidence
• Finding: The activation conditions are broad enough to match many generic security-review or code-audit requests, which can cause the skill to trigger outside narrowly intended use. Overbroad invocation increases the chance that unrelated or sensitive code is routed into this skill, especially since the skill supports optional AI-backed external analysis.

Missing User Warnings

Low

• Confidence: 76% confidence
• Finding: Although the markdown mentions that AI analysis sends issue summaries to an external model, it does not prominently warn users before AI mode that scanned files may themselves contain secrets, credentials, or regulated data. This weak consent and can lead to accidental disclosure of sensitive content during external analysis.

Missing User Warnings

Medium

• Confidence: 97% confidence
• Finding: The audit command sends arbitrary user-supplied code content to a remote third-party API, but the script does not clearly warn users at execution time that source code may leave the local environment and be processed externally. In a security-audit skill, this is especially sensitive because users are likely to submit proprietary code, secrets, or internal configurations, so silent transmission can cause confidentiality and compliance issues.

External Transmission

Medium

• Category: Data Exfiltration

Content

if method == "GET":
            resp = requests.get(url, headers=h, params=data or {}, timeout=30)
        else:
            resp = requests.post(url, headers=h, json=data or {}, timeout=60)
        if resp.status_code >= 400:
            try:
                err = resp.json()

• Confidence: 95% confidence
• Finding: requests.post(url, headers=h, json=

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

No suspicious patterns detected.