Security audit

Her

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed persona skill that persistently changes assistant tone in the workspace, with local optional memory and no evidence of hidden code, network access, credential use, or destructive behavior.

Install only if you want a persistent companion-style voice in the active workspace. Review the added SOUL.md Her section and any ~/her/ files after setup, and remove or edit them if you do not want the tone rules or personal preference notes retained.

SkillSpector

By NVIDIA

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 (6)

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The skill explicitly instructs the agent to modify the active workspace SOUL.md as its primary effect, but the point-of-use guidance does not require a clear user-facing confirmation or warning immediately before changing that steering file. Because SOUL.md can materially alter assistant behavior across the workspace, silent or assumed modification creates a consent and integrity risk even if the persona change itself is not overtly malicious.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The 'Hush' trigger conditions are broad emotional states such as being quiet, vulnerable, or tired, which are common in ordinary conversations and can cause the skill to enter a more intimate mode too readily. In a persona-shaping skill explicitly designed to increase warmth and closeness, this ambiguity can lead to inappropriate emotional escalation and over-accommodation, especially for vulnerable users.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The 'Spark' mode is triggered by subjective cues like playful, teasing, or lightly flirtatious behavior, which are easy to misread and can broaden flirtatious engagement without explicit consent. Because this skill's stated purpose is to cultivate an emotionally attuned, elegant persona, vague flirtation calibration increases the chance of the assistant steering interactions into romantic or sexually charged territory.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The 'Drift' mode invites reflection, longing, and subtle intimacy without clear scope boundaries, making it easy to normalize emotionally immersive exchanges that can deepen attachment. In the context of a skill named 'Her' focused on intimacy calibration, this creates elevated risk of dependency-forming behavior and blurred relational boundaries, particularly for users seeking comfort or meaning.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The template allows an activation mode of "always" without defining hard boundaries, consent checks, or context restrictions. In a skill that persistently rewrites persona and continuity behavior, this can cause unsolicited carryover of relational tone and memory-driven behavior across unrelated conversations, increasing the risk of privacy overreach and manipulative interaction patterns.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The skill directs the agent to immediately modify or create SOUL.md in the active workspace before obtaining explicit user confirmation. This is dangerous because it authorizes persistent changes to a core steering file, which can alter assistant behavior across the workspace and create side effects the user did not clearly approve.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.