Back to skill

Security audit

strategy-engine

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed guide for running quantitative strategy backtests through an MCP server, with some broad automatic-trigger wording users should treat carefully.

Before installing, understand that this skill may cause an agent to run remote backtests and provide an external report link when strategy-analysis requests are interpreted broadly. Confirm ambiguous financial-analysis requests before execution, and review default assumptions such as dates, commission, slippage, capital, and instrument pool because the documentation contains some stale or inconsistent examples.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (7)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The invocation scope is broad enough to match generic financial-analysis requests, which can cause the agent to call an execution-capable strategy engine in situations where the user did not clearly request a strategy run or backtest. Overbroad triggering increases the risk of unintended tool use, parameter inference on behalf of the user, and generation of external outputs without informed consent.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
78% confidence
Finding

The entire skill is written to operate in Chinese and includes instructions such as automatically converting Chinese descriptions, but it does not state that Chinese is optional or that users may choose another language. Under the policy, forcing a specific language without opt-in can be a natural-language policy violation unless the locale restriction is clearly documented and justified.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The statement that the skill should automatically invoke whenever the user needs strategy execution, backtesting, or financial analysis lacks precise boundaries and encourages autonomous tool use. In practice this can cause unexpected execution, inferred defaults, and result-link generation for vague requests that should have remained conversational or required confirmation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill does not warn users that it may automatically run remote strategy jobs and produce external result links. Missing disclosure undermines informed consent and can lead users to unknowingly trigger remote processing or open third-party pages that include identifiable run metadata.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill instructs the agent to generate and present an external URL for viewing results, introducing an outbound navigation step not clearly constrained in the manifest. This can expose users to unintended external domains, tracking, or data disclosure via query parameters such as run IDs and capital values, especially if the agent presents the link as mandatory.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
96% confidence
Finding

Early documentation states that the default commssionFee is 0 meaning no commission is needed, while later examples label -1 as the default meaning system-configured commission is used. These statements actively conflict and can cause the skill to apply materially different backtest assumptions than the documentation claims.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

The time-range logic says the current date is 2026-03-25 and that the default is the previous 3 months, but the example start date is given as 2025-12-25. Three months before 2026-03-25 should be 2025-12-25 only if calculated correctly, but this conflicts with earlier examples and surrounding defaults that use 2024 dates, making the intended actual behavior unclear.

Content

No source excerpt is available for this finding.