Adaptive Reasoning

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:38 PM UTC
• Version:
  2.0.0

Overview

This is a prompt-only skill that adjusts the agent's reasoning depth, with some broad and inconsistent instructions but no evidence of credential access, file access, network activity, persistence, or destructive behavior.

Install this only if you want the agent to automatically manage reasoning depth and possibly use more tokens on complex requests. Users who prefer explicit control should edit or avoid the automatic activation rules, especially the silent activation and unclear threshold rows.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• 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

Findings (4)

Description-Behavior Mismatch

Medium

• Confidence: 92% confidence
• Finding: The skill contradicts itself by claiming no external tools are needed while instructing use of a `session_status` tool and `/reasoning` commands. This can cause an agent framework to attempt unauthorized or unsupported tool/state manipulation, creating unsafe hidden side effects and weakening operator expectations about what the skill is allowed to do.

Intent-Code Divergence

Medium

• Confidence: 94% confidence
• Finding: The documentation says the skill is only mental preprocessing with no external tools, but later includes tool-based and command-based activation guidance. That inconsistency can mislead reviewers and users about the skill's actual behavior, enabling covert control-flow changes or tool usage that bypass expected safeguards.

Vague Triggers

High

• Confidence: 96% confidence
• Finding: Triggering on every user message is overly broad for a preprocessing skill that alters reasoning behavior. A global trigger increases attack surface, can override user intent for quick answers, and may cause persistent hidden mode changes across unrelated requests, especially in sensitive or resource-constrained contexts.

Vague Triggers

Medium

• Confidence: 90% confidence
• Finding: The automatic activation logic is ambiguous because threshold values are corrupted/unspecified and the instructions say 'Do not ask. Just activate.' This creates unconstrained hidden behavior, making it easier for the agent to silently enter elevated reasoning/state modes without clear policy boundaries or user awareness.

VirusTotal

63/63 vendors flagged this skill as clean.