LoR (Loop of Resilience)

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:38 PM UTC
• Version:
  1.0.1

Overview

This skill is a small local verification helper, with broad agent-behavior hooks that users should understand but no evidence of hidden data theft, destructive behavior, or unrelated privileged access.

Before installing, decide whether you want this skill to persist in agent memory and activate automatically on broad complex-task categories. Keep destructive-operation confirmation enabled, and prefer using it explicitly with /lor if you want tighter control.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• 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
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Natural-Language Policy Violations

Medium

• Confidence: 91% confidence
• Finding: The README explicitly says the system forces the agent into a Chain of Thought process, which can encourage collection or disclosure of hidden reasoning that safety-aligned agents should not reveal. In an agent skill context, this is more dangerous because it frames CoT exposure as a feature and may pressure downstream integrations to elicit sensitive internal reasoning rather than just validated outputs.

Vague Triggers

Medium

• Confidence: 90% confidence
• Finding: The skill declares it is 'Auto-triggered on complex tasks' such as coding, architecture, and complex planning, but those conditions are broad and subjective. In an agent environment, this can cause the skill to activate unexpectedly on many sensitive tasks, changing agent behavior and potentially invoking its multi-step reasoning/execution flow without clear user intent.

Natural-Language Policy Violations

Medium

• Confidence: 95% confidence
• Finding: The description explicitly states that the skill implements a Chain of Thought paradigm and automates a 'Raw Reply -> Verification Planning -> Forced Execution -> Result Checking -> Final Output' flow. Mandating hidden reasoning or forced internal processing without user opt-in is risky because it can alter how the agent handles requests, conceal intermediate behavior, and normalize execution-oriented steps that the user did not explicitly authorize.

VirusTotal

64/64 vendors flagged this skill as clean.