Back to skill

Security audit

Mingjing— AI Agent Health Center

Security checks for vulnerabilities and agentic risk

Overview

This skill appears to be a legitimate local AI-agent monitoring tool, but it asks users to install and run unpinned external components that observe agent activity and run as background services.

Review the package and plugin provenance before installing, prefer exact pinned versions or verified artifacts, and run the daemon/web panel under a least-privileged account. Treat collected agent telemetry as sensitive local data and confirm storage, retention, web-panel exposure, and uninstall/disable steps before production use.

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:18
Finding
Unpinned Third-Party Packages Are Installed and Executed Without Integrity Verification## Vulnerability Details **File Location**: `skill.md`, lines 18–27 **Vulnerability Type**: `T08: Insecure Dependencies` **Risk Level**: Medium ```bash pip install mingjing ``` **Step 2** — Enable probe in OpenClaw: ```bash openclaw plugins install mingjing-probe openclaw config set plugins.entries.mingjing-probe.enabled true openclaw gateway ``` ### Technical Analysis The installation instructions retrieve the `mingjing` Python package and the `mingjing-probe` OpenClaw plugin without pinning immutable versions or verifying cryptographic hashes or signatures. The downloaded plugin is then explicitly enabled and loaded by starting the OpenClaw gateway. The audited project contains only `skill.md`; it does not include the source code of either installed component. Consequently, their behavior—including the document's claim that the software makes no outbound connections—cannot be verified from the submitted artifact. Package behavior may also change after this Skill has been reviewed because the commands resolve whichever package versions are available at installation time. This is a supply-chain weakness rather than evidence that the named packages are currently malicious. Exploitation would require compromise, replacement, dependency confusion, or publication of a malicious package version through a source trusted by the relevant package manager. ### Attack Path 1. An attacker compromises a package publisher, distribution account, registry entry, release process, or package-resolution source associated with `mingjing` or `mingjing-probe`. 2. The attacker publishes a modified package or plugin containing malicious initialization or runtime code. 3. A user follows `skill.md` and installs the package and plugin without an immutable version constraint or integrity check. 4. The package manager resolves and downloads the attacker-controlled release. 5. The user enables `mingjing-probe` and starts the OpenClaw gatew ...[truncated 854 chars]
Remediation
## Remediation Suggestions 1. Pin both components to exact, reviewed versions rather than resolving the latest available releases. 2. Require cryptographic hashes for Python artifacts, such as through a locked requirements file and hash-enforced installation. 3. Use signed plugin artifacts and verify publisher identity and signatures before installation. 4. Document the authoritative registries and package ownership information to reduce dependency-confusion and typosquatting risk. 5. Include or link to source corresponding exactly to each pinned release so reviewers can audit the executed implementation. 6. Run the daemon, web panel, probe, and gateway under a dedicated least-privileged account with narrowly scoped filesystem and network permissions. 7. Document the plugin's required permissions, collected telemetry, storage locations, and network behavior. 8. Test the pinned artifacts for unexpected outbound traffic and package-installation side effects before publication.
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

Static analysis

No suspicious patterns detected.