Back to skill

Security audit

Content Engine

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent content-workflow guide, but it asks users to install an unpinned agent dependency and describes scheduled automation that can write into an Obsidian vault without clear review controls.

Install only if you are comfortable with a Chinese-language content automation workflow that may create and update notes in an Obsidian vault. Pin and verify `openclaw-multiagent` before installing it, configure the vault path narrowly, and enable Daily or Weekly hooks only after confirming what sources they scan and where they write results.

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
README.md:14
Finding
Unpinned Third-Party Skill Installation## Vulnerability Details **File Location**: `README.md:14` **Vulnerability Type**: Unverified and unpinned third-party dependency **Risk Level**: Medium **Complete Code Snippet**: ```bash # Install OpenClaw clawdhub install openclaw-multiagent ``` ### Technical Analysis The setup instructions install the third-party `openclaw-multiagent` Skill without specifying an immutable version, cryptographic digest, verified publisher, or trusted source. The dependency is not included in this project, so its implementation and effective behavior could not be reviewed during this audit. Because the installation resolves package content at installation time, a compromised registry account, malicious package update, dependency substitution, or registry compromise could cause users to install content different from what the project author originally intended. Any instructions or executable components supplied by that dependency may subsequently operate with the permissions granted to the Agent environment. ### Attack Path 1. An attacker compromises the dependency publisher, registry entry, or package distribution channel for `openclaw-multiagent`. 2. The attacker publishes a malicious release under the expected package name. 3. A user follows the README and executes the unpinned installation command. 4. The package manager resolves and installs the attacker-controlled release. 5. When the dependency is loaded or invoked, its malicious instructions or code execute with the permissions available to the Agent. 6. Depending on those permissions, the payload could read or modify workspace data, alter Agent behavior, invoke accessible tools, or perform unauthorized external actions. ### Impact Assessment The maximum impact depends on the permissions assigned to OpenClaw and the installed Skill. Potential exposure includes project and workspace files, Agent tool access, workflow integrity, and any credentials available to the Agent pro ...[truncated 189 chars]
Remediation
## Remediation Suggestions 1. Pin `openclaw-multiagent` to a reviewed, immutable version or content digest rather than resolving the latest available release. 2. Document the expected registry, package publisher, and canonical package source to reduce dependency-confusion and typosquatting risks. 3. Verify package signatures or checksums before installation where supported. 4. Audit the dependency's Skill instructions, scripts, hooks, and transitive dependencies before recommending it. 5. Use a lockfile or equivalent integrity manifest if the package manager supports one. 6. Run the dependency with least privilege, restricting filesystem, credential, network, and tool access to what the content workflow requires. 7. Update the installation documentation to include the verified version and integrity-verification procedure.
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 (4)

Natural-Language Policy Violations

Medium
Confidence
88% confidence
Finding
The skill documentation is entirely in Chinese, including setup steps and command descriptions, with no indication that users may choose another language or that the skill is intended only for a Chinese-language or region-specific audience. This creates a natural-language policy concern because it effectively imposes a locale/language requirement without opt-in.

Missing User Warnings

Medium
Confidence
86% confidence
Finding
The skill explicitly describes an automated workflow that creates files in an Obsidian vault and stages content for publication, but it does not include user-facing consent, scope limits, or warnings before modifying local notes or triggering publishing-related steps. In an agent setting, this can lead to unintended filesystem changes or accidental progression toward public distribution of content without the user's informed approval.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The Daily hook is designed to scan external platforms and update the Obsidian topic library automatically, yet the skill provides no privacy notice, source-handling guidance, rate-limit/crawling safeguards, or approval step before ingesting external content into local notes. This creates risk of silent data collection, storage of scraped third-party material, and unintended persistence of sensitive or policy-restricted content in the user's knowledge base.

Natural-Language Policy Violations

Low
Confidence
88% confidence
Finding
This markdown template uses Chinese throughout field values and section headings, which imposes a specific language on users. Under the policy rule, language constraints should either provide user choice or be clearly documented as region- or locale-specific.

Static analysis

No suspicious patterns detected.