Back to skill

Security audit

OpenClaw Memory Kit

Security checks for vulnerabilities and agentic risk

Overview

The skill appears intended to scaffold a sanitized OpenClaw memory workspace, but it enables automatic invocation and asks users to install unpinned local npm dependencies through a referenced script that is not included for review.

Install only if you are comfortable with a persistent OpenClaw memory workspace and review the generated files before use. Prefer explicit invocation, use an isolated target directory, pin and verify npm package versions before installing, and avoid running the missing bootstrap workflow until you can inspect the actual script and dependency commands.

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:14
Finding

Unpinned Third-Party Dependencies Allow Supply-Chain Code Execution

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:14, SKILL.md:54, and references/generated-files.md:46
Vulnerability Type: Unpinned and insufficiently verified npm dependencies
Risk Level: Medium

Vulnerable Code Snippets

SKILL.md:14:

markdown
3. For the current tested OpenClaw build, install `memory-lancedb-pro` together with a matching local `openclaw` package inside the generated state directory. This keeps `openclaw/plugin-sdk` resolvable for the memory plugin on newer releases.

SKILL.md:54:

markdown
- The current tested memory install path is state-local npm dependencies, not `openclaw plugins install memory-lancedb-pro`, because the plugin still expects local `openclaw/plugin-sdk` resolution.

references/generated-files.md:46:

markdown
- `package.json` and `package-lock.json`
  Written when core memory dependencies are installed. The kit uses state-local `openclaw` plus `memory-lancedb-pro` packages so the memory plugin can resolve `openclaw/plugin-sdk` on the current tested OpenClaw release.

Technical Analysis

The workflow directs the Agent to install openclaw and memory-lancedb-pro without specifying exact, reviewed versions, integrity hashes, or a pre-generated trusted lockfile. The phrase “matching local openclaw package” does not impose a deterministic or verifiable version constraint.

Although the generated workspace is expected to contain package-lock.json, the documentation states that this file is written when dependencies are installed. Therefore, the initial resolution and installation occur before the resulting lockfile can provide reproducibility. npm lifecycle scripts may also run during this initial installation.

A mutable package release, compromised maintainer account, malicious transitive dependency, or altered dependency resolution could consequently introduce attacker-controlled code. The referenced bootstrap script is absent ...[truncated 1631 chars]

Remediation
View remediation

Remediation Suggestions

  1. Pin openclaw, memory-lancedb-pro, and all optional plugins to exact reviewed versions rather than ranges or implicit latest releases.
  2. Include a reviewed package-lock.json in the Skill package and use npm ci so installation fails if dependency resolution differs from the lockfile.
  3. Verify registry provenance and package integrity before installation, including expected package owners and integrity hashes.
  4. Use npm ci --ignore-scripts where compatible. If lifecycle scripts are required, document and explicitly approve each required script before allowing it to execute.
  5. Perform dependency installation under a dedicated, least-privileged account without access to unrelated credentials or sensitive directories.
  6. Add automated dependency auditing and provenance checks to the release process.
  7. Include the referenced bootstrap script in the package so reviewers can verify installation commands, registry selection, version constraints, error handling, and lifecycle-script behavior.
  8. Document a controlled upgrade procedure requiring review and lockfile regeneration before dependency versions change.
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/generated-files.md (reported line 7)May include surrounding context.

text
<target-root>/
  .env
  .env.example
  openclaw.json
  package.json

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill enables implicit invocation without any visible trigger constraints, allowing the agent to activate this packaging and sanitization workflow automatically in broader contexts than intended. Because this skill scaffolds and shares memory-system configurations, unintended invocation could expose workspace structure, propagate sensitive configuration details, or cause unreviewed modifications to memory-related assets.

Content

No source excerpt is available for this finding.