T01 · Skill Instruction Hijacking
Error
- Location
skill.md:181- Finding
Persistent Profile-Based Hijacking of Unrelated Agent Responses
- Content
View full analysis
- Remediation
View remediation
Security audit
Cognitive State Tracker - 通用认知状态追踪系统
Security checks for vulnerabilities and agentic risk
Overview
This skill is a personal cognitive-state tracker, but it stores sensitive psychological profiles and uses them to influence all future responses with broad automatic triggers.
Install only if you intentionally want a Chinese-language persistent self-tracking assistant that analyzes journals or logs and uses the resulting profile to shape future replies. Before use, restrict which files or inputs it may analyze, disable or clarify automatic triggers, and ensure you have a way to review, correct, export, and delete stored cognitive-state records.
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 (4)
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 87% confidence
- Finding
The skill name, instructions, trigger descriptions, warnings, and example responses are all written in Chinese, and the document does not offer multilingual operation or ask the user to opt into Chinese output. That creates a natural-language locale constraint without user choice, which matches the language/locale policy violation criteria.
- Content
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 91% confidence
- Finding
The auto-trigger condition 'after_daily_input_detection' is undefined, so the system may activate on content the user did not intend for psychological analysis or long-term storage. Because this skill performs sensitive inference and persistence, ambiguous triggering expands collection scope and can silently capture private material beyond the user's expectations.
- Content
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 94% confidence
- Finding
The skill is designed to ingest diaries, journals, and logs, derive sensitive psychological inferences, and persist them across multiple files, but its overview does not clearly warn users that highly sensitive mental-state profiles and historical records will be continuously created and updated. In a persistent-memory skill handling intimate personal data, omission of explicit disclosure meaningfully undermines informed consent and increases the risk of unexpected retention, overcollection, and privacy harm.
- Content
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 93% confidence
- Finding
The rule to execute upon 'detecting user input (diary/log/review)' lacks clear boundaries, making the trigger scope overly broad for a skill that infers mental state and writes persistent memory. In context, broad detection can cause accidental profiling of unrelated text, work artifacts, or mixed-content notes, increasing both privacy exposure and the chance of storing incorrect or excessively sensitive conclusions.
- Content