Back to skill

Security audit

Opc Os Core

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent one-person-company automation playbook, but it asks agents to run persistent cron workflows, write shared files, use cloud communication channels, publish public content, and potentially republish skills without enough explicit approval and scoping controls.

Review before installing. Enable cron jobs only after setting explicit output directories, notification recipients, cloud-document permissions, retention rules, and approval gates for public posts and SkillHub publishing. Treat monthly republishing and autonomous public posting as requiring human review unless you intentionally delegate that authority.

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
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (7)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

Automatically republishing all skills on a monthly schedule is a materially impactful system change that can propagate bad configurations, vulnerable updates, or unintended behavior at scale without human review. In a skill ecosystem, silent republish automation increases supply-chain and change-management risk because a local compromise or mistaken version diff could be promoted broadly and repeatedly.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill content is written entirely in Chinese and frames the operating model, setup guidance, and governance rules without indicating language flexibility or asking for user preference. In an agent skill, this can override the surrounding user experience and reduce informed use, especially if a user cannot understand safety-relevant instructions, governance constraints, or publishing behavior described by the skill.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The document prescribes persistent shared-file and cloud-document channels for inter-agent communication but does not include any guidance on data classification, access control, retention, or secrets handling. In an AI-native company context, these channels are likely to carry operational, legal, financial, or customer data, so normalizing persistent storage without privacy/security guardrails increases the risk of data leakage, overexposure, and accidental retention.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The reference hard-codes the timezone as Asia/Shanghai and uses Chinese task names and output instructions such as 每日学习 and 300字学习笔记. This appears to force a specific locale/language in natural-language instructions without opt-in or an explicit statement that the skill is intentionally China-specific.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This section defines unattended cron-triggered agent actions that automatically write files into the repository without any explicit approval, safety notice, or guardrails around scope and destination. In an agentic operating-system skill, silent scheduled writes can cause unexpected data modification, overwrite existing artifacts, and normalize autonomous behavior that users may not realize is continuously running.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The weekly inspection job both writes reports and triggers outbound notifications, but the documentation does not warn users that the automation modifies files and communicates externally. In this multi-agent orchestration context, unannounced alerting and report generation can leak operational metadata, create notification spam, or mask autonomous actions as routine background behavior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
72% confidence
Finding

The title includes Chinese text and the only concrete communication example is written in Chinese, with no indication that language choice is optional or tied to a region-specific requirement. That can imply a fixed language/locale expectation without user opt-in, which conflicts with the language/locale policy criteria.

Content

No source excerpt is available for this finding.