Undeclared Tool Scope
Medium
- Category
- MCP Least Privilege
- Confidence
- 70% confidence
- Finding
Without declared permissions the skill's intent is opaque and cannot be validated.
- Content
Security audit
李诞写作心法
Security checks for vulnerabilities and agentic risk
Overview
This is a Chinese writing-style skill with an optional local article checker, and I found no hidden data access, persistence, network behavior, or destructive actions.
Install this if you want Chinese long-form explanatory writing in this specific style. Avoid enabling it as a default for short answers, technical documentation, multilingual work, or tasks where the user’s requested tone and language should take priority.
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
Vulnerability Patterns
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
- MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
- MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (8)
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 88% confidence
- Finding
The skill strongly enforces a Chinese-language writing style without checking the user's requested language, which can cause instruction conflicts, degraded usability, and accidental noncompliance with user intent. While not a classic security exploit, forced locale behavior can become a policy and reliability issue when the agent handles multilingual or accessibility-sensitive requests.
- Content
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The trigger conditions are broad enough to auto-activate on many generic writing, explanation, or teaching requests, which can cause unintended routing and override user-preferred style or safer task-specific behaviors. In an agent system, overbroad activation increases the chance of prompt capture, where this skill silently dominates unrelated requests and changes outputs without clear user consent.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 86% confidence
- Finding
The checklist explicitly requires using '你' to address readers, and later repeats this as a writing requirement. This imposes a specific language/locale style in the skill guidance without any opt-in or alternative for users who may need another language.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 88% confidence
- Finding
The instruction '确保用"你"称呼' is a mandatory language-style constraint. Because the document does not provide an opt-in, audience exception, or locale justification, it conflicts with the policy against forcing a specific language/locale choice.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 93% confidence
- Finding
This code is explicitly framed as a Chinese writing-quality checklist and all user-facing prompts, criteria, and reporting strings are hard-coded in Chinese. The file does not offer any language selection or explain a justified region-specific constraint, which can violate language/locale policy for skills expected to be usable without forced language assumptions.
- Content
Description-Behavior Mismatch
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The manifest describes a writing-method skill for explaining concepts using a seven-step framework while hiding framework titles. In code, the checker additionally requires 2-3 groups of data comparisons, a 4000-6000 character/article length target, and a 10-book tiered reading list, which are substantial output constraints not implied by the manifest description.
- Content
Description-Behavior Mismatch
Low
- Category
- Not specified by scanner
- Confidence
- 80% confidence
- Finding
The module documentation says it validates whether an article conforms to a seven-step framework, but the main execution path groups validation around five core questions and several auxiliary checks. This creates a semantic mismatch between the claimed framework structure and the actual implemented validation logic.
- Content
Static analysis
No suspicious patterns detected.