Back to skill

Security audit

heyyumi

Security checks for vulnerabilities and agentic risk

Overview

HeyYumi provides a coherent venue-search and booking service with disclosed data sharing and explicit confirmation before reservation requests.

Use this for intentional Korean venue searches. Configure credentials outside chat, review booking details before confirming, and expect live-status questions to notify venue staff. Booking contact details go to HeyYumi and the venue; the hosted service's implementation and retention practices were not verified.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Activation includes everyday phrases that do not establish Korean venue-search or reservation intent.

Medium
Category
Trigger Abuse
Confidence
92% confidence
Finding

Triggers such as "오늘 뭐 먹지" (what should I eat today?) and "머리 자르러 갈까" (should I get a haircut?) do not establish that the user wants Korean venue discovery or an external service. They can activate this skill outside its intended scope and steer unrelated conversations toward HeyYumi; however, the required area clarification and explicit confirmation before reservations substantially limit the consequences.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% 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 54)May include surrounding context.

On a headless host the OAuth callback returns to localhost on that host, so a user chatting from a phone usually can't finish it; suggest the API key instead (free at https://heyyumi.ai/mcp, set as HEYYUMI_API_KEY or skills.entries.heyyumi.apiKey).

  • Never ask the user to paste a password, key or token into chat.
bash
# list tools and full argument schemas