Back to skill

Security audit

Ai Intelligent Content Generation

Security checks for vulnerabilities and agentic risk

Overview

The skill is a Chinese AI content-generation helper, but its install instructions run unreviewed code from a mutable remote GitHub repository.

Install only if you are comfortable reviewing or trusting the GitHub repository and its dependencies at the exact time you run the commands. Prefer a pinned commit or signed release, use an isolated virtual environment or container, and do not run it with sensitive credentials or administrator privileges exposed.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (1)

T03 · Remote Payload Retrieval and Execution

Error
Location
SKILL.md:40
Finding

Unpinned Remote Payload Retrieval and Execution

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 40-43
Vulnerability Type: Remote payload retrieval and execution
Risk Level: High

bash
git clone https://github.com/openclaw-skills/ai-intelligent-content-generation
cd ai-intelligent-content-generation
pip install -r requirements.txt
python app.py

Technical Analysis

The installation instructions clone a mutable remote Git repository without specifying a verified commit hash, signed tag, or release artifact. They then install dependencies from the retrieved requirements.txt and execute the retrieved app.py.

Neither requirements.txt nor app.py is included in the audited package, so their contents and security properties cannot be verified. The effective payload may change after this skill has been reviewed. Moreover, Python package installation can execute package build or installation logic before app.py is launched.

This behavior creates a remote code execution channel controlled by the current state of the external repository and its dependency sources. Although no malicious payload is present in the two audited files, the documented procedure crosses the audit boundary and executes unverified remote content.

Attack Path

  1. An attacker compromises the referenced repository, its maintainer account, or a dependency source used by its requirements.txt.
  2. The attacker adds malicious code to app.py, a package build script, or a substituted dependency.
  3. A user follows the installation instructions and clones the repository without pinning a reviewed revision.
  4. pip install -r requirements.txt executes attacker-controlled package installation logic, or python app.py directly launches the modified payload.
  5. The payload runs with the privileges and environment access of the user executing the commands.

Impact Assessment

Successful exploitation permits arbitrary code execution with the invoking user's privile ...[truncated 598 chars]

Remediation
View remediation

Remediation Suggestions

  • Include the executable implementation and dependency manifests within the reviewed skill package so the complete runtime behavior is auditable.
  • Pin the external source to a specific reviewed commit hash rather than cloning the repository's mutable default branch.
  • Prefer signed release artifacts and verify their cryptographic signatures or published checksums before installation.
  • Replace unconstrained dependencies with a lock file containing exact versions and cryptographic hashes, such as a hash-locked requirements file.
  • Install dependencies in an isolated virtual environment or sandbox under a dedicated, least-privileged account.
  • Review all dependencies, including transitive dependencies and package build scripts, before installation.
  • Avoid running the application with administrator privileges and restrict access to credentials, sensitive files, and internal networks.
  • Add an explicit verification step that aborts installation if the repository revision, release signature, or dependency hashes do not match reviewed values.
Vulnerability Patterns
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
83% confidence
Finding

The description is written entirely in Chinese and the tags include 'chinese', which suggests a language-specific skill behavior. There is no visible indication that users can choose another language or that the Chinese-only scope is explicitly justified as region- or audience-specific.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill advertises '内容审核(自动审核)' but does not warn users about how automated moderation/review behaves, including possible false positives, false negatives, bias, or whether content is logged or blocked. This can mislead users into overtrusting the moderation feature and may cause unsafe, noncompliant, or incorrectly filtered content to be published or suppressed.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The manifest description says only 'AI 内容生成助手 - 文章生成、图片生成、视频生成', which broadly covers common content-generation tasks without specifying trigger phrases, activation boundaries, or exclusions. In a manifest file, this kind of general wording can overlap with many ordinary requests and may cause unintended invocation.

Content

No source excerpt is available for this finding.