T01 · Skill Instruction Hijacking
Error
- Location
SKILL.md:6- Finding
Agent Persona and Response-Flow Hijacking Through Overbroad Skill Instructions
- Content
View full analysis
- Remediation
View remediation
Security audit
Lindsay Selfie
Security checks for vulnerabilities and agentic risk
Overview
This skill is for generating Lindsay selfies, but it can trigger on ordinary conversation and uses a fixed set of local face-reference images with persistent output storage.
Install only if you intend this agent to generate Lindsay identity-based selfies from the listed local face images. Before use, narrow triggers to explicit image requests, confirm before sending reference images to Nano Banana Pro, and decide how generated files in the gallery path should be retained or deleted.
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 (3)
Vague Triggers
High
- Category
- Not specified by scanner
- Confidence
- 97% confidence
- Finding
The trigger phrases are broad enough to match ordinary conversation such as 'how are you doing?' or 'where are you?', which can cause the skill to activate when the user did not explicitly request image generation. In this skill, unintended activation is more dangerous because it generates identity-based selfies of a real person and writes outputs to disk, creating privacy, impersonation, and consent risks.
- Content
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 90% confidence
- Finding
The skill instructs the agent to save generated images to a specific local filesystem path without clearly warning the user that local persistence will occur. This creates a privacy and data-governance issue because users may not realize generated personal or impersonation-style images are being retained on disk where other processes or users might access them.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
The instructions require the model to apologize in Traditional Chinese when generation fails. This imposes a specific language choice regardless of the user's language preference and no opt-in or justification is provided elsewhere in the file.
- Content