Back to skill

Security audit

客厅装修避坑指南

Security checks for vulnerabilities and agentic risk

Overview

This is a Chinese home-renovation reference skill with some broader-than-advertised content, but it does not request code execution, credentials, persistence, or system access.

Install this if you want Chinese-language renovation guidance in this publisher's style. Be cautious with demolition, electrical, waterproofing, contract disputes, vendor selection, and home-buying advice: verify important decisions with qualified local professionals, and do not rely on any surfaced WeChat, phone, or off-platform purchase/contact details.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (18)

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest description says this is a 专用 Skill for 客厅装修, implying a focused living-room assistant. However, the referenced content includes kitchen, bathroom, children's room, elderly room, whole-house contracts, budget traps, home inspection, pet-friendly renovation, and legal/rights-defense topics, showing materially broader behavior than the declared scope.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Line L036 instructs the skill to handle 相关空间装修问题 broadly, regardless of project stage, which expands its operational intent beyond the living-room-only purpose stated in the manifest. This is a semantic mismatch between the claimed dedicated scope and the actual documented use behavior.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

Sections such as '选公司与合同避坑' instruct on selecting装修公司, contract traps, payment terms, complaints, and related procurement decisions, which are commercial/transactional advisory functions rather than living-room design guidance. That capability is not justified by the manifest’s stated purpose of helping users design a practical, high-quality living room.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The manifest describes a skill dedicated to living-room renovation guidance, but this reference file contains a full '厨房避坑指南' section with kitchen-specific principles, file mappings, and usage instructions. That is not an implementation detail of living-room advice; it materially broadens the skill’s knowledge domain beyond the claimed客厅-focused purpose.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file includes dedicated behavior specs for '装修翻车急救', general '装修避坑问答', and '收纳储物不踩雷' roles that cover leak repair, contracts, kitchens, bathrooms, children’s rooms, elderly rooms, and buying houses. These instructions define a much broader assistant capability set than a living-room pitfall guide and therefore mismatch the manifest’s claimed narrow specialization.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

This markdown file is entirely written as Chinese-language end-user content and repeatedly presents the material in that single language without any opt-in, alternative locale, or documented region-specific constraint. Under the policy rule for language/locale, forcing one language across the skill content can violate organizational expectations when no user choice or justification is provided.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The manifest describes a skill dedicated to 客厅装修避坑指南, specifically listing living-room-centric topics such as sofa/TV wall/lighting/curtains/balcony/projection/storage and circulation. This reference file includes extensive guidance on bathrooms, kitchens, children's rooms, elderly rooms, contracts, budgets, water/electric systems, buying houses, and whole-home renovation topics, indicating the skill’s actual knowledge scope is much broader than the claimed living-room-only purpose.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill corpus contains embedded contact information, WeChat handles, and shopping/platform guidance despite an instruction model that says such guidance should be banned. This creates a policy-conflict vector where the model may surface external contact or commerce redirection, enabling unwanted solicitation, lead generation, or platform-exit behavior inconsistent with the skill's expected boundaries.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The file is entirely written in Chinese and repeatedly framed around a specific local context such as Xi'an/Shaanxi, without any indication that language or locale is optional or intentionally region-scoped. Under the policy rule, natural-language content that effectively forces a language/locale without user opt-in can be a policy violation.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill is declared as a living-room renovation guide, but the referenced content materially expands into whole-home, kitchen, bathroom, bedroom, legal-rights, and related domains. This scope drift can cause the agent to answer outside its advertised competence and policy boundary, increasing the chance of unsafe, irrelevant, or misleading guidance being presented as if it were in-scope.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

Including legal dispute and rights-enforcement guidance inside a living-room renovation skill creates a hidden capability expansion into legal-adjacent advice. If surfaced to users without clear boundaries or qualification, the agent may provide overconfident legal guidance, mishandle disputes, or bypass expected safeguards for legal content.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Property-selection and investment-style home-buying advice is unrelated to the stated living-room renovation purpose and broadens the skill into financial/real-estate decision support. This can mislead users into relying on the skill for materially different decisions than advertised, raising trust, liability, and policy-boundary risks.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The markdown explicitly instructs the skill to respond using a specific localized voice: “用西安装修课堂‘过日子’的口吻”. This imposes a language/locale-specific style choice on all responses without indicating user choice or opt-in, which matches the policy category for forced language or locale behavior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

This markdown file presents the skill/reference content in a single forced language and does not indicate that the user can choose another language or locale. Under the policy rule for language or locale constraints, forcing one language without user opt-in can be a natural-language policy violation.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The file contains dedicated '二手房避坑类' content on whether to buy old homes, school districts, property age, depreciation, and related real-estate decisions. Real-estate acquisition guidance is a separate advisory capability and is not justified by a manifest focused on living-room renovation workflows.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
96% confidence
Finding

The bundle contains internal answer rules forbidding external-platform/contact guidance while the references repeatedly include WeChat/contact lead-generation content. This creates an instruction-conflict and prompt-injection-style contamination risk where retrieval may surface marketing contacts and cause the agent to violate policy, enabling unwanted solicitation or redirection of users off-platform.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

The section labeled 买房避坑类 discusses choosing high-rise vs 老破小, property carrying costs,拆迁 potential, rescue difficulty, and school-district style purchase tradeoffs. That capability is semantically distinct from advising on 客厅装修避坑 and gives the skill broader real-estate decision support outside its declared purpose.

Content

No source excerpt is available for this finding.