Back to skill

Security audit

unified-world-model

Security checks for vulnerabilities and agentic risk

Overview

This skill is a local, toy world-model helper with disclosed local learning logs and no evidence of network access, credential access, destructive behavior, or hidden execution.

Install only if you are comfortable with a Chinese-language skill that stores local usage counts, errors, notes, and preferences in learned_patterns.json. Do not record secrets or sensitive operational details in learner notes or preferences; delete the JSON file if you want to reset its local memory.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (8)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The documented behavior claims a grounded multimodal world model, but the analyzed implementation reportedly lacks those core functions and instead performs undeclared local persistence and analytics. This mismatch is dangerous because users and orchestrators may trust the skill for simulation and consistency-checking while it silently collects or stores usage data for a different purpose.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The file’s documented and actual behavior is a generic cross-skill telemetry/learning CLI, not a unified multimodal world-model component. This capability mismatch is dangerous because it expands the skill’s effective data-collection scope beyond the declared purpose, undermining user consent, review accuracy, and least-privilege assumptions for a skill that may be trusted for a very different function.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill advertises public availability and describes code-backed behavior, yet declares no explicit tool scope despite static detection of file-write capability. Missing permission boundaries can let the skill write local data without clear user awareness or platform enforcement, increasing the risk of unauthorized persistence or tampering.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill explicitly states that each invocation is recorded by a learner across sessions, including successes, failures, and boundary cases, but provides no user warning or consent flow. Cross-session persistence of interaction-derived data can expose sensitive prompts, operational details, or behavioral profiles beyond the user’s expectation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The module’s docstring, usage text, and user-facing descriptions are entirely in Chinese, and the runtime messages printed to users are also Chinese-only. This imposes a specific language/locale on users without any documented opt-in or alternative, which matches the language-policy violation criteria.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

These functions record arbitrary capability usage, errors, notes, and user preferences into persistent JSON for any supplied skill directory. In context, this creates undocumented telemetry and potential sensitive-data retention across skills, especially because free-form notes and preference values may contain secrets, personal data, or internal operational details.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

This code file contains extensive natural-language instructions and descriptions exclusively in Chinese, including the module summary and CLI help text context, with no indication that users can opt into another language. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale constraint is clearly justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The natural-language description and usage guidance are presented in Chinese only, which can amount to a language policy issue when no user opt-in or locale justification is provided. There is no indication that the skill is intended exclusively for a Chinese-speaking or region-specific context.

Content

No source excerpt is available for this finding.