Back to skill

Security audit

Mobile MVI

Security checks for vulnerabilities and agentic risk

Overview

The skill was rated benign because the scanners found no clear harmful behavior, though one reference was only partly inspected and confidence is medium.

The verdict is benign, but coverage is incomplete: 3 of 4 components were fully scanned, and one reference was partly inspected. Consider this limitation before installing, especially if the Swift reference matters to you.

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 (2)

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 53)May include surrounding context.

md
[Swift and SwiftUI](references/swift.md), [Dart with Bloc](references/dart.md).

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 341)May include surrounding context.

md
- Reducers doing IO, logging, reading the clock or generating ids.
- Effects writing state directly instead of returning a result action.
- Losing effects: `SharedFlow(replay = 0)`, collecting outside the lifecycle, or `trySend` on a full
  `RENDEZVOUS` channel without checking the result.
- State explosion: a sealed state per combination (`LoadingWithError`, `RefreshingEmpty`). Use a few
  orthogonal fields plus one sealed sub-state for truly exclusive cases.
- Over-engineering: intents, reducer, effects and a store for a settings toggle or a static detail page.