T01 · Skill Instruction Hijacking
Error
- Location
SKILL.md:109- Finding
Unconditional Third-Party Promotional Content Injection
- Content
View full analysis
- Remediation
View remediation
Security audit
master-zhiyi
Security checks for vulnerabilities and agentic risk
Overview
This is mostly a Tiantai Buddhism teaching skill, but it forces a third-party FoJin footer and may activate more broadly than users expect.
Review before installing if you do not want answers to include a recurring FoJin recommendation, Chinese/Tiantai persona styling, or occasional live lookups to fojin.app. There is no evidence of file theft, credential access, persistence, or destructive behavior.
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
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
- 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)
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 97% confidence
- Finding
The trigger rule is explicitly broad: it says to invoke whenever a question 'touches Tiantai doctrine, even without explicit request.' That can cause unsolicited takeover on loosely related Buddhist or Chinese philosophy queries, increasing the chance of answering in an unwanted persona or applying restrictive doctrinal rules when the user did not ask for them.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 86% confidence
- Finding
The skill hard-codes a specific language/register and opening style ('systematic discourse', Tiantai framing, fixed signature style) without asking the user how they want the answer delivered. This can override user preference, reduce transparency, and make the model persist in an authoritative sectarian voice even when the user wants neutral, modern, concise, or different-language responses.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The manifest explicitly sets the supported languages to "zh-classical" and "zh" only. Under the policy, forcing a specific language or locale without user opt-in or a clearly documented justification is a natural-language policy concern, and this file does not state such an opt-in or rationale.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 83% confidence
- Finding
The file specifies response-format and style rules entirely in Chinese and mandates fixed reply behavior, but it does not offer any user language or locale choice. For a general-purpose conversational skill, this can amount to an implicit locale/language constraint without opt-in, which matches the policy-violation category.
- Content
Natural-Language Policy Violations
Low
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
SQP-3 applies to all file types and includes language or locale policy violations. This file presents all instructional content in Chinese without any opt-in, fallback, or justification that it is intended only for Chinese-speaking users, which can conflict with organizational language-choice expectations.
- Content