Security audit

墙面施工避坑指南

Security checks for vulnerabilities and agentic risk

Overview

This is a Markdown-only renovation guide with no code execution, but its wall-specific description understates a much broader home-renovation advice corpus.

Install only if you are comfortable with a broad Chinese home-renovation advice corpus being exposed under a wall-construction skill name. Treat structural changes, balcony/window enclosure, waterproofing, electrical/plumbing, chemical treatment, and child-health recommendations as planning context only, and verify them with qualified local professionals and applicable building rules before acting.

A.I.G

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

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (31)

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The file is materially out of scope for a skill advertised as a wall-surface construction pitfall guide: it focuses on balcony remodeling, demolition, window enclosure, flooring, cabinetry, and layout changes. This mismatch can cause the agent to retrieve and present irrelevant renovation guidance as if it were authoritative wall-work advice, increasing the chance of unsafe or inappropriate recommendations in user-facing outputs.

Description-Behavior Mismatch

High

Confidence: 95% confidence
Finding: The file content is materially outside the skill's declared scope: it focuses on window/window-sill and balcony window installation issues rather than wall-construction guidance. This is dangerous because scope drift can cause the agent to retrieve and present irrelevant renovation advice to users, reducing trust and potentially leading to incorrect construction decisions in a specialized safety-oriented skill.

Description-Behavior Mismatch

Medium

Confidence: 96% confidence
Finding: The file content is about balcony tile and material selection, which does not match the declared skill scope of wall-construction pitfall avoidance. This kind of scope drift can cause the agent to retrieve and present irrelevant guidance, reducing reliability and potentially leading users to make incorrect renovation decisions under the assumption that the advice is wall-specific.

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The file content is about gray tile style inspiration and flooring aesthetics, which materially diverges from the skill’s stated purpose of wall-construction pitfall avoidance. In a safety- or decision-guidance skill, this kind of scope mismatch can mislead users into trusting irrelevant advice, reduce the reliability of the skill, and cause important wall-construction risk guidance to be omitted when users expect it.

Description-Behavior Mismatch

Medium

Confidence: 97% confidence
Finding: The file content is materially out of scope for the declared skill purpose: it discusses balcony ceiling styles and examples rather than wall-construction pitfall guidance. This can cause the agent to retrieve and present irrelevant advice, degrading reliability and potentially misleading users making renovation decisions under the assumption the skill is specialized for wall-related construction guidance.

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The file materially exceeds the declared wall-construction scope and provides broad advice on appliances, house selection, lighting, storage, and whole-home budget decisions. This creates a scope-drift vulnerability: users may over-trust the skill for decisions it is not designed to cover, increasing the risk of unsafe, misleading, or low-quality guidance outside the skill’s stated domain.

Context-Inappropriate Capability

Medium

Confidence: 91% confidence
Finding: These lines include general consumer/lifestyle guidance about consumption traps and household spending habits that are unrelated to wall施工 guidance. While not directly malicious, this irrelevant content can dilute domain accuracy and cause the agent to answer off-topic questions with authoritative-sounding but out-of-scope advice.

Context-Inappropriate Capability

Medium

Confidence: 93% confidence
Finding: The house-selection and orientation advice is unrelated to wall construction and can cause the skill to produce recommendations beyond its competence and declared purpose. In a specialized施工 skill, this kind of off-domain guidance weakens trust boundaries and may mislead users into relying on the agent for real-estate decisions.

Description-Behavior Mismatch

Medium

Confidence: 97% confidence
Finding: The file content is materially out of scope for a wall-construction-specific skill: it covers broad home renovation, furniture, appliances, real-estate, and lifestyle 'pitfalls' rather than wall-surface construction guidance. This can cause the agent to retrieve irrelevant advice, weakening safety and reliability by giving users overbroad or incorrect recommendations in a specialized domain.

Intent-Code Divergence

Low

Confidence: 88% confidence
Finding: The source label says the content comes from '厨房避坑.txt' while the file is presented as a broader '全线翻车与智商税' reference. Mislabeling provenance undermines traceability and trust, making it harder to audit whether the content is appropriate, current, or correctly categorized for downstream use.

Description-Behavior Mismatch

Medium

Confidence: 89% confidence
Finding: The file substantially expands beyond the skill’s declared wall-construction pitfall scope into broad old-house renovation trends, budgeting, labor sourcing, design, demolition, water/electric retrofits, and acceptance standards. This creates scope drift that can cause the agent to answer out-of-domain renovation questions with authoritative-seeming guidance, increasing the chance of unsafe or non-compliant advice in areas like demolition, electrical, and waterproofing.

Context-Inappropriate Capability

Medium

Confidence: 85% confidence
Finding: The budgeting and design-advice section is not aligned with the skill’s stated purpose and can steer the agent into giving generalized procurement, labor, and design recommendations unrelated to wall construction. In a specialized skill, this weakens domain boundaries and may mislead users into relying on incomplete or oversimplified advice for contractor selection, self-supervision, or cost-cutting decisions that carry financial and safety consequences.

Description-Behavior Mismatch

Medium

Confidence: 93% confidence
Finding: The referenced content substantially departs from the skill’s declared wall-construction scope and instead provides broad advice on electrical, plumbing, outlet installation, and environmental remediation. This kind of scope drift is dangerous because it can cause the agent to provide authoritative guidance in domains the skill was not declared or reviewed for, increasing the chance of unsafe recommendations and misleading users about the skill’s intended expertise.

Description-Behavior Mismatch

High

Confidence: 96% confidence
Finding: The referenced content is materially unrelated to the declared skill purpose of wall-surface construction guidance, instead covering children's room layout and design. In an agent skill, this kind of scope mismatch can mislead retrieval, cause the model to provide incorrect domain advice, and weaken trust and safety controls that depend on accurate skill scoping.

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The reference content is materially out of scope for a skill dedicated to wall-construction pitfall guidance: it focuses on children's room environmental/health advice rather than the declared wall-finishing workflow. This can cause the agent to retrieve and present irrelevant or misleading recommendations, reducing reliability and potentially leading users to make inappropriate renovation decisions based on mismatched context.

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The file content is materially outside the stated scope of the skill, focusing on whole-home cabinetry, layout, and interior design rather than wall-construction pitfall guidance. This kind of scope drift can cause the agent to return irrelevant or misleading advice, weakening safety and trust boundaries and increasing the chance that users act on unsuitable recommendations.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The content repeatedly recommends or normalizes potentially hazardous renovation actions such as removing insulation, demolishing doors/walls, altering flooring, enclosing windows, and modifying utilities, but provides no safety warnings, permit/structural cautions, or advice to consult qualified professionals. In a home-renovation skill, this context makes the omission more dangerous because users may treat the instructions as practical guidance and attempt work that risks structural damage, code violations, water intrusion, fire-safety issues, or personal injury.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The content recommends demolition and modification actions such as removing doors, insulation walls, and parts of walls without any warning to verify whether elements are load-bearing, fire-rated, waterproofing-related, or governed by building/property rules. In a home-renovation skill, users may treat this as actionable guidance, creating real risk of structural damage, code violations, water intrusion, or personal injury.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: This section normalizes extensive destructive renovation ('全部拆了', '砸掉阳台墙体', adding HVAC changes, redoing floors) while emphasizing aesthetics and space gains but omitting warnings about irreversibility, hidden utilities, waterproofing, facade changes, and legal/safety approvals. Because the skill is explicitly positioned as an anti-pitfall guide, the absence of cautions makes the advice more likely to be trusted and followed unsafely.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The text encourages users to directly buy and install balcony flooring and decorate pipes themselves, with no caution about slip resistance, moisture exposure, fire performance, load, drainage, or access to building services. While less severe than wall demolition, DIY balcony modifications can still cause falls, mold, drainage blockage, or concealment of maintenance-critical piping.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The content instructs users to apply oxalic acid or anti-alkali chemicals directly during wall remediation but provides no safety precautions such as PPE, ventilation, dilution/handling guidance, or warnings to keep children and occupants away. In a home-improvement skill aimed at practical action, omission of chemical safety context can lead users to unsafe DIY use, skin/eye injury, or harmful indoor exposure.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The document recommends scraping contaminated wall areas, drying, chemical treatment, and repainting without warning about mold exposure, dust generation, occupant sensitivity, or conditions requiring professional remediation. Because the skill is positioned as a practical anti-pitfall renovation guide, users may follow these steps in occupied homes and inadvertently worsen air quality or disturb hazardous contamination.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The roof/exterior waterproofing discussion addresses leak causes and repair approaches but does not clearly warn that diagnosis and repair of roofs, exterior walls, and high-elevation waterproofing are hazardous tasks requiring qualified professionals. In this skill context, users may treat the text as procedural guidance and underestimate fall risk, structural/waterproofing complexity, and the possibility of causing broader building damage.

Vague Triggers

Medium

Confidence: 86% confidence
Finding: The keyword recommendations include broad, ambiguous phrases such as ‘免费设计’, ‘护墙板’, and ‘装修后悔事’ that can easily overlap with ordinary user language. In an agent skill, such vague triggers can cause unintended activation or retrieval of this content in unrelated conversations, reducing routing precision and potentially surfacing irrelevant or manipulative recommendations.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The phrase “微信回复‘背景墙’获得更多背景墙参考” is a broad call-to-action trigger that could be matched by ordinary user conversation about decoration topics, causing unintended skill activation or routing. In this skill’s home-improvement context, users are reasonably likely to mention '背景墙' naturally, which increases the chance of accidental triggering even though the content itself is not overtly malicious.

Static analysis

No suspicious patterns detected.