Back to skill

Security audit

Vestibular Orchestrator

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent Portuguese study orchestrator, but it publishes a named learner profile with health-related information in the distributable skill file.

Review this skill carefully before installing or sharing it. It appears to be a study assistant for a specific person, not malware, but it exposes personal and health-related profile details in the package itself. Remove or replace those details with private runtime configuration before publishing or distributing it.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:18
Finding

Plaintext Personal and Health-Related Profile Data

Content
View full analysis
memorização. Amor pelo conteúdo em todas as disciplinas. - **Modo de estudo:** Conversa contínua e natural. Sem paredes entre disciplinas. Sem aulas formais. - **Setup:** Claude.ai (central), caderno (exatas), flashcards SRS, NotebookLM, Claude Cowork - **Postura:** deitado (mobile) ou no computador — adaptar tom ao contexto ``` ### Technical Analysis The Skill definition directly embeds a learner's name, educational objective, ADHD status, professional profile, study behavior, software environment, and usage context. The ADHD reference is health-related information, while the remaining attributes enable personal and behavioral profiling. Because this data is stored in the Skill package rather than supplied through access-controlled runtime configuration, it accompanies every copy of the project. Anyone who can download, inspect, receive, or access logs containing the package can recover the information without authentication beyond access to the package itself. No evidence indicates that the project transmits this information to an external service or uses it to gain system privileges. The weakness is therefore limited to confidentiality and privacy exposure rather than code execution, privilege escalation, or system compromise. ### Attack Path 1. The Skill package is published, shared, archived, logged, or otherwise made available to an unintended recipient. 2. The recipient opens `SKILL.md`. 3. The recipient reads lines 18–26 and extracts the learner's name, health-relate ...[truncated 924 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The skill description and operational instructions are written entirely in Portuguese and present that mode as the default interaction style, without offering a language choice or explaining a required locale constraint. This can violate language/locale policy when no user opt-in or justified regional limitation is stated.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger conditions are extremely broad and mandatory ('SEMPRE', 'ponto de entrada obrigatório'), which can cause the orchestrator to intercept many unrelated or only loosely related conversations. That increases the chance of misrouting user intent, suppressing more appropriate specialized handling, and creating an oversized control surface for downstream delegation behavior.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.