Back to skill

Security audit

BEE Belief Extraction Engine

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent installer for a persistent OpenClaw memory plugin, with privacy and supply-chain considerations users should understand before enabling it.

Install only if you want OpenClaw agents to build long-lived memory from sessions. Consider pinning a reviewed package version, reviewing the npm/GitHub package before global installation, disabling extraction when handling sensitive conversations, and periodically inspecting or deleting the SQLite belief database.

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:33
Finding
Unpinned Global Installation of a Third-Party Extension## Vulnerability Details **File Location**: `SKILL.md`, lines 33–41 **Vulnerability Type**: Unpinned third-party dependency installed globally from mutable sources **Risk Level**: Medium ### Vulnerable Code ```markdown **From npm (recommended):** ```bash npm install -g @skysphere-labs/openclaw-bee ``` **From GitHub (latest):** ```bash npm install -g github:skysphere-labs/openclaw-bee ``` ``` ### Technical Analysis The installation instructions retrieve and globally install third-party code without pinning an audited package version or immutable Git commit. The npm command resolves the package version associated with the current distribution tag, while the GitHub command retrieves the repository's current default revision. Both sources can change after this Skill has been reviewed. The project contains only `SKILL.md`; it does not include the extension's source, a lockfile, an integrity hash, a signature, or reproducible-build information. Consequently, the effective package payload, npm lifecycle scripts, and runtime extension behavior cannot be verified from the audited artifact. Global npm installation also increases exposure because package installation scripts execute with the installing user's privileges and place executable content in a shared user or system-level npm location. This finding does not establish that the referenced package is malicious; it identifies an unsafe, mutable supply-chain trust boundary. ### Attack Path 1. An attacker compromises the npm publisher account, package distribution channel, or referenced GitHub repository. 2. The attacker publishes malicious package content or modifies the repository's default revision. 3. A user follows the documented unpinned global installation command. 4. npm retrieves the attacker-controlled version and may execute its lifecycle scripts with the installing user's privileges. 5. The user restarts the OpenClaw gateway as instructed, loading the ...[truncated 939 chars]
Remediation
## Remediation Suggestions 1. Pin the npm dependency to a specifically reviewed version instead of using the mutable latest release: ```bash npm install -g @skysphere-labs/openclaw-bee@<reviewed-version> ``` 2. If installation from GitHub is necessary, pin it to a full immutable commit SHA: ```bash npm install -g github:skysphere-labs/openclaw-bee#<full-reviewed-commit-sha> ``` 3. Publish and verify package integrity hashes or cryptographic signatures before installation. 4. Review the package contents and npm lifecycle scripts before executing the installation. 5. Prefer a project-local or otherwise isolated installation over a global installation where supported. 6. Run the extension and OpenClaw gateway under a dedicated, least-privileged operating-system account with restricted filesystem and credential access. 7. Provide a lockfile, software bill of materials, source reference, and reproducible-build instructions so the reviewed source can be matched to the installed artifact. 8. Consider initially installing with lifecycle scripts disabled for inspection, then enabling only scripts that have been explicitly reviewed.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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)

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The skill explicitly states that it extracts beliefs via a lightweight LLM call and stores them indefinitely in a SQLite database, but it does not present a prominent user warning about external transmission of session content or persistent retention. This creates a real privacy and data-governance risk because users may invoke the skill without understanding that potentially sensitive session data will be sent to a third-party model provider and retained across sessions.

Static analysis

No suspicious patterns detected.