Security audit

Shadow Market

Security checks for vulnerabilities and agentic risk

Overview

This skill is a small local prototype for scoring prediction gaps, with no network access or credential use, though it can write local JSONL output when its script is used.

Install only if you are comfortable with a prototype that records prediction-capture data locally when its Python script is used. Avoid entering confidential event names or business forecasts unless you control the output path and retention.

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

MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 70% confidence
Finding: Without declared permissions the skill's intent is opaque and cannot be validated.

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The manifest and module/class documentation present this as a 'prediction market' that 'trades' and produces 'shadow prices'. In practice, the code just stores submitted probabilities in memory, calculates a derived score from the shallow/deep spread, and appends JSON lines to a local file; there is no order book, position handling, settlement, matching, or actual market mechanism.

Intent-Code Divergence

Low

Confidence: 98% confidence
Finding: The inline comment says 'Shadow price = spread × depth_gap_factor × time_decay', suggesting time affects pricing. However, the calculation only uses spread and a depth-based visibility factor; timestamps are collected on predictions but never incorporated, so the comment materially misstates how prices are computed.

Missing User Warnings

Low

Confidence: 87% confidence
Finding: The code appends event and prediction-derived market data to a local file automatically, which can persist potentially sensitive or user-supplied information without notice or consent. In this skill's context, event names may encode proprietary research topics, forecasts, or internal business signals, so silent disk writes can create unintended data exposure and retention risks.

Static analysis

No suspicious patterns detected.