Pydaqua SpaceAutonomySkill

Security checks across malware telemetry and agentic risk

Overview

The skill appears to be a local computational terrain-classification demo, but its space-navigation framing should not be treated as real autonomous navigation capability.

Install only if you want a toy or research-style terrain classification helper. Do not rely on it for real spacecraft, robotics, rover, drone, vehicle, or other safety-critical navigation decisions without independent validation, explicit guardrails, and human review.

SkillSpector

By NVIDIA

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (2)

Description-Behavior Mismatch

Medium

Confidence: 93% confidence
Finding: The implementation does not provide autonomous navigation; it performs a small heuristic similarity check against three hard-coded terrain signatures and emits a text decision. In a safety-critical space autonomy context, this capability overstatement is dangerous because downstream operators or orchestration systems could trust it for real navigation decisions despite lacking robust sensing, validation, path planning, or safety controls.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The skill exposes a very generic `navigate` command without any visible scope restrictions, parameter constraints, or invocation boundaries. In an autonomous navigation context, an overly broad trigger increases the chance of unintended activation or misuse by other agents/components, which could lead to unsafe decisions or unauthorized execution paths.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal