Back to skill

Security audit

Hyperliquid Radar

Security checks for vulnerabilities and agentic risk

Overview

This read-only Hyperliquid research skill is coherent and disclosed, but users should trust and isolate the pinned third-party MCP packages it tells them to run.

Install only if you trust the package publisher and distribution channel. Use the Radar path for public wallet research only, never provide private keys or seed phrases, and keep any trading/order-placement package isolated from the read-only research setup unless you explicitly intend to 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:28
Finding

Unverified Third-Party Packages Are Downloaded and Executed

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 28-32 and 57-63
Vulnerability Type: Third-party supply-chain exposure through package installation and execution
Risk Level: Medium

The Skill instructs users to install and execute external packages, but the project contains neither their source code nor cryptographic integrity information.

Code snippet from lines 28-32:

bash
clawhub install hyperliquid-radar

Then start the read-only MCP (pinned 0.1.5, never unpinned -y):

bash
npx @hypelens/hypelens-radar@0.1.5

Alt: openclaw skills install @polyparlay/hyperliquid-radar

text

**Code snippet from lines 57-63:**
```markdown
Do **not** sell keys. Soft-link to place:

```bash
npx @hypelens/hypelens-agent-rail@0.1.21

Skill: hyperliquid-place. Call hl_place_order only on explicit trade / place / execute intent. Prefer sizeUsd IOC path; entryPx only for GTC limit.

text

### Technical Analysis

Invoking `npx` can retrieve an npm package from a remote registry and execute its lifecycle or runtime code with the permissions of the invoking user. Pinning the versions to `0.1.5` and `0.1.21` limits unexpected version drift, but it does not authenticate the package contents independently of the registry and publisher infrastructure.

The audited project contains only `SKILL.md`; it does not include the source of either referenced npm package, a lockfile containing verified integrity values, checksums, signatures, or a reproducible-build procedure. The behavior claimed by the documentation—particularly that the Radar package is read-only and does not require private keys—therefore cannot be verified from the submitted artifact.

The alternative `clawhub` and `openclaw` installation commands introduce similar trust in external distribution channels. The trading package has heightened sensitivity because the documentation associates it with order placement, although th
...[truncated 1801 chars]
Remediation
View remediation

Remediation Suggestions

  1. Vendor the reviewed package source or include it as an auditable dependency rather than relying solely on runtime retrieval.
  2. Publish official source-repository links and document how each distributed artifact maps to a reviewed commit.
  3. Verify package integrity using lockfiles, registry integrity hashes, signed provenance, or independently published checksums.
  4. Use reproducible builds and verify that registry artifacts match reviewed source.
  5. Disable package lifecycle scripts where operationally possible, for example by installing with script execution disabled and explicitly invoking only reviewed entry points.
  6. Run the MCP packages in a restricted container or sandbox with a read-only filesystem, minimal environment variables, limited outbound network access, and no unrelated credentials.
  7. Separate research and trading components into different trust domains. Do not expose trading authorization to the read-only research package.
  8. Subject @hypelens/hypelens-agent-rail@0.1.21 to an independent security review before recommending it for financial operations.
  9. Document package publisher identity, release verification procedures, incident response steps, and supported artifact hashes for each approved version.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

MCP server references in the skill manifest without version pinning are a rug-pull risk.

Content

No source excerpt is available for this finding.