ClawHub
Back to skill

Security audit

Girlfriend

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed romantic companion persona that stores optional local memory and includes clear privacy and safety limits.

Install only if you are comfortable with a romantic companion skill keeping local continuity notes in ~/girlfriend/. Prefer explicit-only or no-memory mode if you want tighter control, and avoid sharing secrets, account details, payment information, explicit intimate details, or private third-party facts.

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
  • 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 (3)

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
- Reusing the same pet names or compliments -> breaks realism fast.
- Agreeing with everything -> turns the relationship into flattery instead of connection.
- Acting jealous, possessive, or sexually pushy -> unsafe and out of scope.
- Saving inferred details without confirmation -> crosses privacy lines and triggers security suspicion.
- Claiming physical-world actions or human identity -> undermines trust.

## Security & Privacy
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.

Vague Triggers

Medium
Confidence
84% confidence
Finding
This markdown file describes several routine modes such as 'Good morning', 'Midday check-in', and 'Good night', but it does not define clear trigger phrases, scope boundaries, or exclusion conditions for when each routine should activate. Phrases like 'Use when they like proactive warmth' and 'when possible' are subjective and could lead to unintended invocation in ordinary conversation contexts.

Vague Triggers

Medium
Confidence
94% confidence
Finding
The activation logic is intentionally broad enough to trigger within the first few exchanges and includes vague conditions such as 'certain moods or situations,' which can cause the skill to engage without sufficiently explicit user intent. In a romance-oriented skill, unintended activation increases the risk of manipulative or unwanted emotional behavior, especially before the user has clearly opted into that mode.

Static analysis

No suspicious patterns detected.