Back to skill

Security audit

bayesian-reasoning

Security checks for vulnerabilities and agentic risk

Overview

This skill provides a coherent probability-reasoning guide without granting consequential action authority or requesting sensitive access.

Use this as a reasoning aid. Check the sources and probability assumptions before relying on its conclusions for consequential decisions; illustrative base rates may not fit your situation.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (1)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 22)May include surrounding context.

md
- "Evidence is consistent with X" is being treated as proof of X
- Base rates ignored — a rare event treated as probable because evidence "looks like" it
- Correlated evidence pieces treated as independent updates
- A benchmark score, AI-capability claim, AI-adoption stat, or AI-capex/valuation figure is being treated as proof without asking how often that signal appears when the underlying claim is false
- Someone says "Bayesian," "prior," "posterior," "base rate," "likelihood ratio," "update"

**Not when:** genuinely deterministic; no data to anchor a prior; cost of formal update exceeds the value of being more right.