Construction Mediation Kg

• Outcome: Review
• Latest audit: May 28, 2026 · 8:22 AM UTC
• Version: 1.0.0

Overview

This skill is not malicious, but it handles sensitive legal mediation data with local storage and optional third-party AI calls that are not tightly scoped or consent-gated.

Review before installing for real matters. Use the local knowledge-graph parts freely for generic analysis, but do not send confidential case facts, personal data, settlement positions, or privileged materials to the AI-backed functions unless your organization approves the provider and data handling. If using case or evidence management, store the generated JSON files in a protected directory and define your own deletion and retention process.

SkillSpector

Vulnerability Patterns

• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 86% confidence
• Finding: The skill advertises no declared permissions, yet its documented behavior includes environment-variable access and file read/write operations. This creates an authority mismatch: a reviewer or runtime policy may treat the skill as low-risk while it can actually access secrets and persist case data, increasing the chance of unauthorized data exposure or ungoverned state changes.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 94% confidence
• Finding: The manifest describes a knowledge-graph mediation assistant, but the file also documents operational case management, evidence handling, local persistence, dashboarding, and outbound AI-provider calls. This mismatch is dangerous because users and platform controls may authorize the skill for advisory analysis while it performs materially broader actions on sensitive legal data, including storage and external transmission.

Description-Behavior Mismatch

Medium

• Confidence: 88% confidence
• Finding: The documentation expands from decision-support into operational case, evidence, and dashboard management, implying storage, tracking, and processing of sensitive dispute information beyond the declared scope. In a legal-mediation context this is more dangerous because the data likely contains confidential party, payment, evidence, and procedural records, so silent scope expansion raises privacy and governance risks.

Description-Behavior Mismatch

Medium

• Confidence: 91% confidence
• Finding: The optional AI integration adds external legal-search and case-analysis services, with API keys sourced from environment variables and case content potentially sent to third-party providers, but this is not reflected in the manifest description. For a construction-dispute mediation skill, that creates significant confidentiality risk because sensitive legal facts, evidence summaries, and negotiation positions could be transmitted off-platform without clear disclosure or consent.

Description-Behavior Mismatch

Medium

• Confidence: 95% confidence
• Finding: This module sends user-supplied case descriptions, legal issues, and case_info to the external Alibaba DashScope API. In a legal mediation context, those inputs can contain highly sensitive or confidential dispute details, so undisclosed or overly broad off-platform transmission creates a real data-exposure and privacy risk, especially if the skill is presented as a local knowledge-graph expert rather than an external AI relay.

Static analysis

No suspicious patterns detected.