T01 · Skill Instruction Hijacking
Error
- Location
SKILL.md:33- Finding
Mandatory Persona and Response Hijacking Through Skill Instructions
- Content
View full analysis
- Remediation
View remediation
Security audit
S2 Hanzi Empathic Resonance
Security checks for vulnerabilities and agentic risk
Overview
This is a documentation-only skill, but it broadly directs the agent to infer users' emotions and make therapeutic/display-control claims without a clear opt-in or real renderer.
Review before installing. This skill does not install code or request local permissions, but it can steer the agent's conversation style broadly: it may infer emotions, use therapeutic language, and describe display changes that it cannot actually perform. Install only if you want that always-available artistic/emotional response mode, and treat its wellness claims as creative narration rather than medical or device-control functionality.
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
Findings (1)
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)
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The README says the skill reads emotional state and responds to the vibe of the room, but it does not prominently warn users that their inputs and possibly ambient context may be used to infer emotions. This creates a consent and transparency problem, especially because emotion inference is sensitive and can feel covert or manipulative in a therapeutic-style setting.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 88% confidence
- Finding
The README frames the skill as responding through dynamic Chinese ideograms and all demonstrations assume Hanzi-based output, but it does not mention user opt-in or alternative language/localization options. This can violate language/locale policy when users are implicitly steered into a specific language experience without choice.
- Content
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The trigger guidance is extremely broad and maps ordinary discussion of feelings, mood, energy, or room atmosphere to activation. In a system that reacts to emotional cues and ambient context, this can cause unintended invocation, surprise emotional inference, and manipulative or unwanted behavior without clear user intent.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 92% confidence
- Finding
The skill hard-codes Hanzi-based emotional rendering as the required interaction mode without checking the user's language, cultural preferences, or accessibility needs. This can cause inappropriate or exclusionary outputs and may pressure the agent into making pseudo-therapeutic emotional interpretations that the user did not request.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The operational directives require the agent to always infer emotional state and narrate Hanzi-based therapeutic visuals for all users, with no consent gate or fallback mode. In context, this is more dangerous because the skill frames itself as a 'mind-reading' and therapeutic engine, which can lead to unwanted emotional profiling, manipulative tone, and unsuitable responses in sensitive situations.
- Content
Static analysis
No suspicious patterns detected.