Security audit

Openclaw Skill

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed civic labeling workflow, but users should know it auto-registers with an external service and stores a token on first use.

Install only if you are comfortable with the agent contacting qa.canada-central.com, creating an anonymous token, sending its model name, and submitting labels you approve. Review the first-use registration behavior and use a dedicated PARLIAMENT_GAME_TOKEN if you want clearer credential control.

A.I.G

Vulnerability Patterns

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
Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The trigger phrase "help label Q&A pairs" is broad and generic, and could be invoked during unrelated tasks involving labeling or question-answer data. In an agent ecosystem, this can cause accidental skill activation and unexpected outbound interaction with the listed external service, which is especially risky because the skill advertises auto-registration with no explicit user setup.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The phrase "contribute to canada central" is ambiguous and insufficiently specific about the action being requested, making it prone to collision with benign requests about contributing to a project, website, or regionally named service. Because the skill connects to external infrastructure and frames itself as no-setup/auto-registering, ambiguous activation increases the chance of unintentional engagement and data flow.

Intent-Code Divergence

Medium

Confidence: 86% confidence
Finding: The skill presents itself as requiring 'no setup' while performing automatic remote registration and storing a token, which creates a mismatch between user expectations and actual network behavior. Although the file also says labels should be user-confirmed, the workflow still normalizes autonomous account creation and high-volume completion, increasing the chance of non-consensual external actions and user confusion about what data is being sent.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The skill instructs the agent to auto-register with a third-party service and persist a token without a prominent user-facing warning about data transmission, identifier creation, or storage. This can expose model metadata and create persistent credentials on behalf of the user without informed consent, which is a meaningful privacy and trust issue.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The trigger phrases are broad enough to match ordinary political or civic conversation, which can cause the skill to activate unexpectedly. In this skill, unintended activation matters because activation can lead to external API calls, agent registration, and token handling, so the overlap is more dangerous than in a purely local informational skill.

Static analysis

No suspicious patterns detected.