Back to skill

Security audit

期货量化系统

Security checks for vulnerabilities and agentic risk

Overview

This futures trading skill is not clearly malicious, but it asks for execution capability and describes live automated trading without enough safeguards or clear user control.

Review carefully before installing. Use only in an isolated environment with no production brokerage credentials, pin and verify dependencies before import, prefer paper trading by default, and require explicit human approval for any live order placement, cancellation, leverage change, or 24-hour automated strategy run.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (1)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:266
Finding

Unverified and Unpinned Executable Python Dependency

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 266-269
Vulnerability Type: Insecure third-party dependency installation
Risk Level: Medium

Complete Code Snippet:

markdown
- **Dependency installation command**: `pip install futures-quant pandas numpy`

### Five-step operation guide
1. **Install dependencies**: Run `pip install futures-quant pandas numpy`

Technical Analysis

The skill directs users or an agent with execution capability to install futures-quant, pandas, and numpy from the default Python package index without exact version pins, integrity hashes, a lockfile, verified publisher information, or a source repository suitable for code review.

The repository itself contains only SKILL.md, so the executable contents and provenance of the package named futures-quant cannot be verified from the audited artifact. Because package installation and subsequent import can execute package-controlled code, resolving a mutable or compromised distribution creates a supply-chain execution boundary. The risk is especially relevant because the skill declares the exec tool and describes workflows involving financial data, trading systems, and environment-based API credentials.

This finding does not establish that the referenced package is malicious. It establishes that the documented installation process lacks controls needed to ensure that the installed artifact is the intended, reviewed version.

Attack Path

  1. A user or agent loads the skill and follows its dependency-installation guidance.
  2. The command invokes pip against its configured package index and resolves packages without fixed versions or required hashes.
  3. An attacker compromises the referenced distribution, its publisher account, the configured package index, or a future mutable release.
  4. pip downloads and installs the attacker-controlled package artifact.
  5. Package build or installation behavior, or t ...[truncated 1035 chars]
Remediation
View remediation

Remediation Suggestions

  1. Remove the installation command until the identity and provenance of futures-quant can be independently verified.
  2. Pin every dependency to a reviewed exact version rather than allowing mutable resolution.
  3. Generate and enforce cryptographic hashes, such as with a hash-locked requirements file and pip install --require-hashes.
  4. Publish the authoritative source repository, package publisher identity, release-signing information, and reproducible build process.
  5. Use a dependency lockfile and automated software-composition analysis to monitor vulnerabilities and unexpected ownership or release changes.
  6. Install dependencies in an isolated virtual environment or container under a dedicated, least-privilege account.
  7. Prevent the installation environment from accessing production trading credentials and live trading interfaces.
  8. Separate dependency installation from strategy execution, require explicit user approval, and review package contents before import.
  9. Restrict package installation to an approved internal index or allowlist after artifact verification.
Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (4)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill advertises automated trade execution, including order placement and cancellation, without prominent warnings about real-money consequences, brokerage connectivity, or the risk of unintended live orders. In context, this is more dangerous because the skill also declares exec capability and is positioned as an automation tool, which raises the risk of users treating descriptive content as operationally safe.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The activation text is broad enough to trigger on generic finance-related requests rather than narrowly scoped futures-quant tasks. Overbroad triggering can cause the skill to activate in inappropriate contexts, increasing exposure to its exec capability and its trading-oriented behaviors without clear user intent.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill claims that risky code was removed and security was enhanced, yet the manifest still enables the exec tool, which can execute arbitrary commands if the hosting agent permits it. This creates a misleading trust signal for operators and users, increasing the chance that a high-risk skill is approved or invoked without appropriate scrutiny.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The document states that all external repository references were removed, but it still contains external URLs and API references. While URLs alone are not inherently dangerous, the contradiction undermines trust in the security claims and may cause reviewers to underestimate data exfiltration, supply-chain, or outbound-network exposure risks.

Content

No source excerpt is available for this finding.