Back to skill

Security audit

AC Buying Consultant

Security checks for vulnerabilities and agentic risk

Overview

This is a plain instruction-only AC buying guide with no hidden tools, credentials, persistence, or unsafe execution behavior.

Review the sizing examples and local efficiency/refrigerant claims before relying on the recommendations for a purchase, especially because AC standards and prices vary by region and year. From a security perspective, this skill is low risk because it is instruction-only and does not ask the agent to run commands, access accounts, or store data.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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 (3)

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The example calculation is internally inconsistent with the sizing formula given earlier: 168 sq ft × 25 BTU should not yield 4,200 BTU, and the example then overrides the computed result with a larger recommendation. In a consumer-buying skill, contradictory sizing guidance can cause users to mistrust or misapply the method, potentially leading to incorrect AC selection, wasted money, or poor cooling performance.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill first states that it must never recommend an undersized AC, but the example later instructs the agent to provide specs for a smaller unit if the user insists. That contradiction weakens the safety guardrail in a domain where bad recommendations can cause financial loss, continuous overuse, underperformance, and accelerated equipment wear.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The manifest description uses British English spelling ('prioritisation'), and the document consistently assumes a fixed wording/style without explicitly offering the user a language or locale preference. Under the policy rule for language/locale constraints, skills should avoid forcing a specific locale unless the user opts in or the constraint is justified.

Content

No source excerpt is available for this finding.