Back to skill

Security audit

Actually - prediction markets

Security checks for vulnerabilities and agentic risk

Overview

The skill is mainly a prediction-market lookup tool, but it can become a real-money trading tool when configured with a Polymarket private key.

Use this as read-only unless you deliberately want agent-assisted trading. Do not set `POLYMARKET_PRIVATE_KEY` casually; if you do, pin and review the npm package version, isolate the MCP process, use a limited wallet, set strict order and daily caps, and require manual confirmation for every trade.

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:38
Finding

Unpinned npm Package Is Downloaded and Executed Through npx

Content
View full analysis
Remediation
View remediation
"] } } } ``` 2. Install dependencies ahead of execution using a committed lockfile containing package integrity metadata. Use a deterministic installation mechanism such as `npm ci`. 3. Configure runtime execution so it fails if the package is not already installed, preventing `npx` from silently downloading mutable code. 4. Verify package provenance, signatures or attestations where available, and review the selected package version and its transitive dependencies before deployment. 5. Run the MCP server in a sandbox or isolated container with minimal filesystem and network access. 6. Provide only the environment variables required for read-only operation. Do not expose `POLYMARKET_PRIVATE_KEY` unless the operator has explicitly enabled trading. 7. For trading deployments, use a restricted signer or wallet with strict balance, transaction, per-order, and daily spending limits. Separate read-only and trading server profiles. 8. Monitor and approve dependency updates rather than accepting new releases automatically. ]]>
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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 (2)

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
94% confidence
Finding

Financial trading is materially riskier than passive market lookup, yet the skill's stated purpose is only to check prediction market prices. Embedding trading capability in this context makes prompt-injection and user-confusion attacks more dangerous because a tool selected for information retrieval may also be able to spend funds if the private key is configured.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill is presented as a read-only market lookup tool, but the documented behavior includes order placement and redemption when an environment variable is present. This creates a capability mismatch that can lead operators or upstream agents to invoke the skill in a higher-risk financial transaction context than intended, increasing the chance of unauthorized or socially engineered trades.

Content

No source excerpt is available for this finding.