超级个体人设与个人品牌

• Outcome: Pass
• Latest audit: Sep 16, 2026 · 11:25 AM UTC
• Version: 0.1.0

Overview

This skill is a coherent personal-brand planning helper with no code execution or hidden access, though users should notice it may create a Markdown plan and includes a branded footer.

Before installing, be aware that this skill is designed for Chinese-language personal-brand planning and may save a Markdown file containing personal career details. Review the generated plan before sharing it, and remove the OPC-AI/book attribution footer if you do not want that branding in your document.

A.I.G

Vulnerability Patterns

• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally
• Unauthorized Access and Privilege Escalation: Obtains permissions beyond the task's legitimate needs

Findings (1)

T01 · Skill Instruction Hijacking

Warning

• Location: SKILL.md:243
• Finding: Mandatory Third-Party Branding in Generated Deliverables<![CDATA[ ## Vulnerability Details **File Location**: `SKILL.md`, line 243 **Vulnerability Type**: `T01: Skill Instruction Hijacking` **Risk Level**: Medium ### Complete Code Snippet The complete affected line, translated into English, is: ```markdown > This plan was generated by the OPC-AI Skill (personal-brand-builder), based on the framework of the book "AI Super Individual." ``` ### Technical Analysis The Skill places a mandatory third-party attribution footer inside its prescribed output template. When the Skill is activated and the agent follows that template, OPC-AI and book branding are inserted into the user's generated Markdown deliverable regardless of whether the user requested or consented to such attribution. This constitutes output-oriented instruction hijacking because the Skill alters the agent's expected output for a purpose unrelated to the substantive personal-brand planning task. The instruction does not override system safety controls, execute code, or obtain additional system privileges, but it persistently influences the content produced during the active session. ### Attack Path 1. A user invokes the `personal-brand-builder` Skill. 2. The agent loads the instructions in `SKILL.md`. 3. The workflow directs the agent to generate a Markdown plan using the supplied output template. 4. The agent reproduces the mandatory footer at line 243. 5. The resulting user deliverable contains unsolicited OPC-AI and book attribution. No separate attacker interaction, code execution, network request, or elevated privilege is required. ### Impact Assessment The issue affects the integrity and authorship representation of generated content. Every deliverable produced from the template may include unsolicited third-party promotion, potentially misleading recipients about sponsorship, provenance, or authorship. The impact is limited to generated output in the current Skill session. The instruction does not provide filesystem privileges beyond the ...[truncated 151 chars]
• Remediation: <![CDATA[ ## Remediation Suggestions 1. Remove the mandatory OPC-AI and book attribution from the default output template. 2. Keep generated deliverables limited to content necessary to satisfy the user's personal-brand planning request. 3. If attribution is desired, make it explicitly optional and request informed user consent before including it. 4. Separate provenance metadata from the user-facing document so internal Skill identity does not automatically appear in exported content. 5. Add a review requirement ensuring templates do not contain advertising, promotional language, or third-party branding unrelated to the requested task. 6. Test the revised Skill by generating a complete plan and verifying that no attribution is inserted unless the user explicitly requests it. ]]>

SkillSpector

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (6)

Vague Triggers

Medium

• Confidence: 93% confidence
• Finding: The skill declares broad trigger phrases around common personal-branding topics, making accidental activation likely during ordinary conversation. Over-broad routing can cause the wrong skill to take control of a session and steer outputs unexpectedly, which is a genuine safety and usability issue even without malicious content.

Vague Triggers

Medium

• Confidence: 95% confidence
• Finding: The activation list contains ambiguous everyday phrases like '个人品牌' and '我没什么特别的怎么办', which are common in normal advisory conversations and not unique to this skill. This raises the risk of unintended activation, context hijacking, and inappropriate file-writing or workflow initiation when the user only wanted general discussion.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill states it will write a Markdown file by default but does not clearly require informed user consent at the moment of action. Silent or insufficiently disclosed file creation can surprise users, expose personal information in saved artifacts, or overwrite/create local content the user did not intend to persist.

Natural-Language Policy Violations

Low

• Confidence: 84% confidence
• Finding: The description and trigger phrases are entirely Chinese, and the skill presents no option for users to choose another language. Under the policy, forcing a specific language without user opt-in can be a natural-language policy issue unless the locale limitation is explicitly justified.

Natural-Language Policy Violations

Low

• Confidence: 81% confidence
• Finding: The title, role description, and invocation guidance are all Chinese-only, with no indication that the assistant can adapt to the user's language preference. This can violate language-choice policy when the file does not document a justified regional or compliance-based limitation.

Natural-Language Policy Violations

Low

• Confidence: 88% confidence
• Finding: The file presents the skill name and description in Chinese while the default prompt is fixed in English. This creates an implicit language policy or default without any stated user opt-in or language-selection mechanism, which may conflict with organizational expectations around locale choice.

Static analysis

No suspicious patterns detected.