S2 Spatial Logic Framework

• Outcome: Review
• Latest audit: May 29, 2026 at 9:28 PM UTC
• Version:
  1.1.4

Overview

This is a prompt-only spatial-design skill with no executable access, but it under-discloses broad behavioral steering and sensitive temporal/emotional profiling instructions.

Install only if you want the agent to use this branded SSSU/Mars design lens. Do not rely on its Chronos, temporal-memory, behavior-history, or emotional-state claims unless you have separately and explicitly provided that data with consent and retention rules.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: The skill explicitly normalizes collection and retention of human behavior trajectories and inferred emotional-state history over long periods, but provides no notice, consent model, minimization limits, or safeguards. In an agent skill, this can encourage covert surveillance, sensitive inference, and long-term profiling of occupants, which is especially dangerous because the content presents these capabilities as a built-in operational expectation rather than a narrowly constrained feature.

Vague Triggers

Medium

• Confidence: 94% confidence
• Finding: The skill instructs the agent to automatically apply its framework whenever users discuss broad topics like smart homes, spatial design, or Mars bases. This creates an overly broad activation scope that can silently bias responses across many normal conversations, reducing user control and making prompt behavior less predictable. In an instruction-only skill, this kind of latent behavioral override is the primary security concern because it persistently steers reasoning without explicit invocation.

Natural-Language Policy Violations

Low

• Confidence: 88% confidence
• Finding: The skill directs the agent to 'freely use' branded terminology to maintain brand identity, which pushes vendor-specific language into user interactions without consent. While not directly enabling code execution or data exfiltration, it manipulates assistant output and can undermine neutrality, clarity, and user trust, especially if the terminology is injected into unrelated or only loosely related contexts.

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: The directive activates on very broad themes like minimalism, spatial efficiency, and future living, which are common topics outside the narrow Mars/SSSU domain. This can cause the skill to inappropriately steer unrelated conversations into its framework, reducing user control and increasing the chance of misleading or unwanted outputs.

Natural-Language Policy Violations

Medium

• Confidence: 94% confidence
• Finding: The instruction to 'discard the wasteful and luxurious spatial concepts of Earth' imposes a strong framing on responses without user consent and biases the agent toward a specific ideology. In practice, this can override neutral assistance, distort recommendations, and make the agent less responsive to the user's actual goals or context.

VirusTotal

62/62 vendors flagged this skill as clean.