Back to skill

Security audit

三省六部

Security checks for vulnerabilities and agentic risk

Overview

This skill is a multi-agent orchestration package with expected dashboard, audit, and model-provider behavior, but its defaults and install path create review-worthy exposure and supply-chain risk.

Review before installing. Use a virtual environment or container, pin dependencies and the edict runtime to a reviewed version or commit, change dashboard binding to 127.0.0.1 unless you intentionally expose it, add authentication before any non-local access, and decide whether long-retention audit logs are appropriate for your data and compliance requirements.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (2)

T08 · Insecure Dependencies

Error
Location
scripts/install.sh:12
Finding

Unpinned Third-Party Dependencies and Externally Supplied Runtime Code

Content
View full analysis
/dev/null || pip install -q --break-system-packages pyyaml requests flask prometheus-client 2>/dev/null || echo "Dependency installation skipped" ``` The startup script also installs dependencies dynamically: ```bash python3 -c "import yaml" 2>/dev/null || { echo "📦 Installing dependencies..." pip install -q pyyaml requests flask } ``` The documentation instructs users to retrieve an unpinned external runtime: ```bash pip install edict-openclaw git clone https://github.com/cft0808/edict.git cd edict pip install -e . ``` ### Technical Analysis The project does not include the Python `edict` implementation imported by its startup script and tests. Its effective runtime therefore depends on code obtained separately from a package registry, a mutable Git repository, or a pre-existing installation. Neither the package installations nor the Git installation are pinned to reviewed, immutable artifacts. The dependency declarations use only lower bounds, such as `pyyaml>=5.4`, and the shell scripts install packages without exact versions or cryptographic hashes. Consequently, future package releases or changes to the repository's default branch can alter the code executed after this artifact has been reviewed. The installer also falls back to `pip --break-system-packages`. This can alter the host's system-managed Python environment rather than containing dependencies in a dedicated virtual environment, increasing the scope of a compromised or incompatible dependency. Because the actual server implementation is absent, its runtime ...[truncated 1446 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
scripts/install.sh:25
Finding

Dashboard Exposed on All Network Interfaces Without Authentication in the Generated Default Configuration

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (11)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The natural-language content of the README is presented entirely in Chinese, while installation commands and identifiers suggest a broader audience. Under the language/locale policy, forcing a single language without user opt-in or documented justification can be a policy violation.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding

The skill documents shell and network-capable operations such as pip install, git clone, docker-compose, kubectl, exposed dashboard launch, and service startup, but it does not declare any tool scope or permission boundaries. In an agent environment, that mismatch can cause overbroad execution authority, making it easier for an invoking system or operator to run impactful commands without explicit consent or least-privilege controls.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This section includes audit logging and user-data access examples, including personal information, actor identifiers, IP address, and long-term retention, but provides no explicit privacy warning, consent requirements, or data-minimization guidance. That omission can normalize collection and storage of sensitive data in ways that violate privacy expectations or legal requirements when copied into real deployments.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

The example enables auto_deploy: True, which promotes autonomous release behavior from a multi-agent orchestration system into deployment workflows. While not malicious by itself, automatic deployment without an explicit human approval gate can amplify mistakes, unsafe model actions, or compromised pipeline inputs into immediate production changes.

Content

Scanner excerpt · SKILL.md (reported line 545)May include surrounding context.

md
ci_cd={
        "pipeline": "gitlab-ci",
        "stages": ["build", "test", "security_scan", "deploy"],
        "auto_deploy": True
    }
)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill describes production deployment and launching a dashboard on 0.0.0.0 without a clear warning about exposure, authentication hardening, network restrictions, rollback safety, or change-control requirements. In this context, users may treat the examples as safe defaults and expose administrative surfaces or automate deployment into sensitive environments with insufficient safeguards.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
82% confidence
Finding

The file is written entirely in Chinese for descriptions, comments, panel names, and alert labels, with no indication that users can choose another language or that the configuration is intended only for a Chinese-language deployment. Under the policy for natural-language issues, a locale-specific presentation without opt-in or documented justification can be a language/locale policy violation.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · assets/config-example.yaml (reported line 100)May include surrounding context.

yaml
providers:
    openai:
      api_key: "${OPENAI_API_KEY}"
      base_url: "https://api.openai.com/v1"
      
    anthropic:
      api_key: "${ANTHROPIC_API_KEY}"

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · scripts/install.sh (reported line 17)May include surrounding context.

sh
pip install -q pyyaml requests flask prometheus-client 2>/dev/null || pip install -q --break-system-packages pyyaml requests flask prometheus-client 2>/dev/null || echo "⚠️ 依赖安装跳过,请手动安装: pip install pyyaml requests flask prometheus-client"

# 创建配置目录
mkdir -p ~/.edict/config
mkdir -p ~/.edict/logs
mkdir -p ~/.edict/data

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
86% confidence
Finding

This markdown file shows a quick-start example that enables audit=True and launches a dashboard, which may collect, display, or retain user/system activity data. The README does not include any user-facing warning about privacy, data retention, or system-impact considerations for these behaviors.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

This shell script contains user-facing natural-language strings entirely in Chinese, including the installation banner and usage instructions. Under the policy, forcing a specific language without offering a language or locale choice is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

The manifest description is written as '三省六部制 · OpenClaw Multi-Agent Orchestration System', which introduces a Chinese-language framing without any indication that users can opt into another language or that the skill is specifically intended for a Chinese locale. Under the stated policy, forcing a specific language or locale without user choice can be a natural-language policy issue.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.