Security audit

Lembrete Agua

Security checks for vulnerabilities and agentic risk

Overview

This is a simple hydration reminder skill with some over-specific Goiânia assumptions, but it does not contain hidden execution, credential access, exfiltration, or privileged behavior.

Install only if you want a Portuguese hydration reminder centered on Goiânia-style climate assumptions. Confirm your weight, schedule, and water entries explicitly, and do not treat the generated intake target as medical advice if you have health conditions or fluid restrictions.

A.I.G

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

SkillSpector

Vulnerability Patterns

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
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Natural-Language Policy Violations

Medium

Confidence: 91% confidence
Finding: The skill hard-codes Goiânia-specific context in its description without checking the user's location or preferences, which can cause unnecessary localization assumptions and implicit handling of location-sensitive context. While not directly exploitable in a severe way, it can mislead users and normalize use of inferred locale data without opt-in.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The activation phrases are broad enough that ordinary discussion about hydration could unintentionally trigger the skill, causing unsolicited behavior and possible collection or processing of personal health-related input. In a conversational agent, ambiguous invocation boundaries can lead to confused-deputy style interactions where the user did not intend to enter tracking mode.

Natural-Language Policy Violations

Medium

Confidence: 88% confidence
Finding: The reminder content repeatedly assumes the user is in Goiânia and references local weather conditions without opt-in, which can result in inaccurate or privacy-sensitive personalization. This is a design flaw around contextual assumptions rather than a direct exploit, but it can still reduce trust and cause inappropriate recommendations.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The consumption logging triggers are vague and overlap with normal speech, so the skill may record water intake when the user is merely talking about drinking rather than reporting an action. This can corrupt health-tracking data and create unintended state changes without clear user intent.

Natural-Language Policy Violations

Medium

Confidence: 90% confidence
Finding: The climate adaptation logic is tied to Goiânia-specific conditions and seasonal assumptions without offering locale selection, which can produce inaccurate health guidance and location-based personalization without consent. In a health-adjacent tracking skill, incorrect environmental assumptions may lead to over- or under-recommendation of intake targets.

Static analysis

No suspicious patterns detected.