Back to skill

Security audit

Thần Cửu (Jaskies)

Security checks for vulnerabilities and agentic risk

Overview

This is a Vietnamese personality-analysis skill with no code execution, but it asks for legal-name and birth-date details and instructs agents to save all generated analyses persistently without clear consent, retention, or deletion controls.

Install only if you are comfortable with the agent saving Thần Cửu reports that may contain full names, dates of birth, relationship context, and personality inferences. Prefer asking the agent not to save results unless you explicitly want retention, avoid entering unnecessary identity-document details, and delete knowledge/ThanCuu/ records when no longer needed.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:47
Finding
Unnecessary Persistent Storage of Personal and Inferred Data## Vulnerability Details **File Location**: `SKILL.md`, lines 47–50 **Vulnerability Type**: Unprotected persistent storage of sensitive personal data **Risk Level**: Medium ### Vulnerable Code Snippet ```markdown ## Lưu ý quan trọng - **Độ chính xác:** Luôn sử dụng đúng tên và ngày sinh trên giấy tờ tùy thân. - **Nguyên tắc cốt lõi:** Tuyệt đối không áp dụng các khái niệm số Master (11, 22, 33) của Thần số học truyền thống vào đây. Hệ thống này chỉ vận hành trên 9 con số cơ bản. - **Lưu trữ:** Mọi kết quả phân tích sẽ được lưu vào thư mục tri thức `knowledge/ThanCuu/` để anh có thể tra cứu lại bất cứ lúc nào. ``` Related input guidance appears in `SKILL.md`, lines 34–41, and `references/GioiThieuThanCuu.md`, lines 80–83 and 149–152. It directs users toward names and birth dates taken from identity documents. ### Technical Analysis The Skill instructs the Agent to save every generated analysis to the persistent `knowledge/ThanCuu/` directory. These analyses are derived from legal names and dates of birth and may include relationship details and inferred personality or psychological characteristics. Persistent storage is not required to perform the stated calculation. The instruction does not require informed consent before storage and provides no data-minimization, pseudonymization, access-control, encryption, retention, expiry, or deletion requirements. Consequently, information needed only transiently can remain accessible beyond the originating session. This is classified as `T09: Insecure Skill Coding Practices` because the configuration directs plaintext persistence of sensitive personal and inferred data without appropriate safeguards. The reviewed evidence does not establish attacker-controlled instruction persistence, so it is not classified as Agent Memory Poisoning. ### Attack Path 1. A user provides a legal name and date of birth for an individual or two individuals. 2. The Agent calculates the requested indicators and generates personality o ...[truncated 1011 chars]
Remediation
## Remediation Suggestions 1. Remove mandatory persistence and process names and birth dates transiently by default. 2. Ask for explicit user consent before saving any analysis, clearly identifying what data will be stored, where it will be stored, and for how long. 3. Store only the minimum necessary information. Exclude identity-document numbers, document images, addresses, and full birth dates unless strictly required. 4. Pseudonymize or redact legal identifiers from retained reports where possible. 5. Apply restrictive filesystem permissions so unrelated users, Skills, and processes cannot read the records. 6. Encrypt retained sensitive data at rest using appropriately managed keys. 7. Define automatic expiration and secure deletion procedures, and provide a user-accessible deletion mechanism. 8. Avoid instructing users to submit identity-document images; a manually entered name and birth date are sufficient for the declared function. 9. Document whether generated results may enter backups, synchronization systems, logs, or other secondary storage.
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 (15)

Natural-Language Policy Violations

Medium
Confidence
84% confidence
Finding
The usage instructions address the user only in Vietnamese and prescribe Vietnamese trigger phrasing, with no indication that other languages are supported or that Vietnamese is required for a justified region-specific purpose. This can constitute a language policy issue because the skill appears to enforce a locale implicitly without opt-in.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill explicitly states that all analysis results will be stored in `knowledge/ThanCuu/`, while the inputs include personal data such as full names and dates of birth. Failing to clearly warn users about persistent storage and retention of this personal data creates a privacy risk and may lead to collection or reuse of sensitive data without informed consent.

Ssd 3

Medium
Confidence
97% confidence
Finding
The instruction that all analysis results will be stored for later lookup implies retention of user-supplied personal data, including names, birth dates, and inferred personality analysis. In this skill context, the retained output may also contain sensitive derived profiling data, which increases privacy and misuse risk if stored without clear consent, access controls, or retention limits.

Natural-Language Policy Violations

Medium
Confidence
87% confidence
Finding
The skill mandates a single 'most common' Latin transliteration for non-Latin names without asking the user or preserving ambiguity. Because transliteration is not unique across languages and conventions, this can silently alter inputs and produce inconsistent or biased personality outputs, especially when the method treats the transformed spelling as analytically significant.

Natural-Language Policy Violations

Medium
Confidence
92% confidence
Finding
This markdown file presents all user-facing content exclusively in Vietnamese, and there is no indication that the skill is region-specific or that users can opt into this language. Under the policy rule for natural-language violations, forcing a specific language without user choice is a reportable issue.

Natural-Language Policy Violations

Medium
Confidence
96% confidence
Finding
This file consists entirely of Vietnamese natural-language content and does not indicate that the user can opt into this language or select another locale. Under the policy, forcing a specific language without user choice or clear region-specific justification is a natural-language policy violation.

Natural-Language Policy Violations

Medium
Confidence
97% confidence
Finding
The entire skill content is written in Vietnamese and does not offer any language or locale choice. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale restriction is clearly documented and justified, which is not present here.

Natural-Language Policy Violations

Medium
Confidence
95% confidence
Finding
This file presents all user-facing natural-language content in Vietnamese and does not indicate that users can opt into another language. Under the policy criteria, forcing a specific language without user choice is a natural-language policy violation unless the locale restriction is explicitly justified.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The document explicitly asks users to provide personal information 'theo CCCD hoặc CMTND' (government ID-based identity details) but provides no notice about data minimization, purpose limitation, storage, consent, or handling safeguards. In the context of a personality-analysis skill, collecting identity-linked personal data is unnecessary for core functionality and increases privacy, re-identification, and misuse risk if the information is stored, logged, or shared.

Natural-Language Policy Violations

Low
Confidence
95% confidence
Finding
The entire skill content is written in Vietnamese and does not indicate any option for the user to choose another language or locale. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation.

Natural-Language Policy Violations

Low
Confidence
89% confidence
Finding
The file contains only Vietnamese-language text and points to Vietnamese content, with no indication that the user can choose another language or opt in to this locale. Under the stated policy, forcing a specific language without user choice is a natural-language policy concern.

Natural-Language Policy Violations

Low
Confidence
89% confidence
Finding
This file presents all user-facing content in a single language and does not indicate that Vietnamese is an intentional, region-specific requirement. Under the language/locale policy, forcing one language without opt-in can be a natural-language policy violation.

Natural-Language Policy Violations

Low
Confidence
79% confidence
Finding
This markdown file consists entirely of Vietnamese natural-language content and provides no indication that the skill is region-specific or that users can opt into this locale. Under the policy rule for language and locale, forcing a specific language without user choice or a justified scope is a natural-language policy concern.

Natural-Language Policy Violations

Low
Confidence
83% confidence
Finding
A language or locale policy issue can arise when a skill effectively forces a single language without user opt-in or justification. This file presents all instructions and questions only in Vietnamese and does not indicate that the skill is region- or language-specific or provide an alternative language option.

Natural-Language Policy Violations

Low
Confidence
92% confidence
Finding
This markdown file presents all user-facing content exclusively in Vietnamese and does not offer an alternative language or indicate that the skill is region- or language-specific. Under the natural-language policy rule, forcing a specific language without user opt-in can be a locale-policy violation.

Static analysis

No suspicious patterns detected.