ClawHub
Back to skill

Security audit

Abductive Reasoning

Security checks across malware telemetry and agentic risk

Overview

This is a markdown-only reasoning skill that teaches structured hypothesis generation and does not request system access or persistence.

Use this skill as a structured way to generate and test hypotheses. Do not use intuitive abductive judgments about real people as a basis for accusation, confrontation, medical action, financial action, or disciplinary action without corroborating evidence and an appropriate process.

SkillSpector

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The example presents successful accusation of a thief based largely on intuitive, non-articulable impressions, then frames the confession as validating the method. In a skill meant to teach abductive reasoning, that can encourage readers to apply intuition-driven suspicion to real people without adequate safeguards, increasing risk of false accusation, bias, or harmful confrontation.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.