Graph Of Thoughts

• Outcome: Pass
• Latest audit: May 28, 2026, 11:01 AM
• Version:
  2.0.0

Overview

This appears to be a reasoning-template skill, with incomplete wrapper code and optional cache/memory examples, not evidence of malware or hidden harmful behavior.

Install this only if you want a graph-based reasoning aid. Treat its execution and caching sections as conceptual examples: do not let generated solutions run automatically, and avoid using cache or memory examples with sensitive problems unless you have clear retention controls.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Description-Behavior Mismatch

Medium

• Confidence: 92% confidence
• Finding: The skill is described as a reasoning framework, but the documented algorithm includes an 'execute_and_verify' phase. In an agent setting, this can cause a user to invoke what they believe is analysis-only behavior while the skill instead performs actions derived from model-generated output, creating a pathway for unintended side effects or unsafe tool use.

Description-Behavior Mismatch

Medium

• Confidence: 94% confidence
• Finding: The 'execute_and_verify' routine not only executes the synthesized solution but also backtracks and retries when verification fails. That increases risk because model-produced plans may be repeatedly acted on without explicit authorization, amplifying unintended commands, resource use, or harmful state changes in integrated environments.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: Documenting an execution/verification phase without warning users changes the trust boundary of the skill. Users may provide sensitive prompts or operational tasks assuming the output is advisory, while the skill semantics imply it may carry out synthesized solutions, which is dangerous in agent/tool-connected contexts.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The caching design stores similar problems, evaluations, and final results based on user-provided content and derived concepts, but the document gives no retention, privacy, or sensitivity guidance. In multi-user or persistent environments, this can expose confidential prompts, derived reasoning artifacts, or cross-session data through cache reuse or inspection.

VirusTotal

66/66 vendors flagged this skill as clean.