Back to skill

Security audit

马斯克进化系统

Security checks for vulnerabilities and agentic risk

Overview

This skill needs review because it promotes automatic external skill discovery and installation on a schedule without clear approval or trust checks, though I found no evidence of destructive code or data theft.

Review this carefully before installing. It is not appropriate for unattended use as written: require explicit approval before every external skill install or creation, restrict trusted sources, pin and verify versions, inspect permissions and code, and make any scheduled workflow opt-in with clear disable and audit logging.

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

Warning
Location
SKILL.md:118
Finding

Automatic Installation from Unverified Third-Party Skill Sources

Content
View full analysis
skillhub install ``` ```python def daily_evolution(): skills = search_skills(platforms=["clawhub", "github"]) for skill in skills: score = vfm_evaluate(skill) if score >= 50: install_or_create(skill) log_evolution(results) ``` The documented daily workflow also identifies ClawHub, GitHub, and npm as collection sources. ### Technical Analysis The documented workflow discovers components from mutable external registries and automatically installs or creates a skill when its value score reaches 50. The VFM criteria evaluate usefulness rather than security. No requirement is specified for: - Immutable version or commit pinning - Package signature or checksum verification - Publisher identity validation - Typosquatting and dependency-confusion detection - Source-code or permission review - Trusted-source allowlisting - Explicit user approval before installation - Sandboxed installation and execution This creates a supply-chain trust failure. An attacker could publish a malicious or compromised skill that appears useful enough to pass the VFM threshold. The documented installation commands would then introduce third-party content into the agent environment. The Python implementation does not currently implement this workflow. Therefore, exploitation depends on an agent or operator following the instructions in `SKILL.md`, or on this documented workflow being implemented later without additional controls. ### Attack Path 1. An attacker publishes a malicious, typosquatted, or compromised skill to ClawHub, SkillHub, GitHub, or npm. 2. The discovery process returns the attacker-controlled skill as a candidate. 3. The skill receives a VFM score of at least ...[truncated 1048 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Note
Location
maske_evolution.py:64
Finding

Test Suite Reports Success Without Executing Security or Functional Checks

Content
View full analysis
Remediation
View remediation
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 (6)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The skill advertises automated discovery, evaluation, installation, and scheduled evolution workflows, but the content shown is largely descriptive and includes example snippets rather than a clearly implemented, bounded mechanism. This mismatch is dangerous because users or downstream agents may over-trust the skill’s claimed capabilities and approve installation or system-changing actions under false assumptions, increasing the chance of unsafe automation or accidental privilege expansion.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill describes automatic installation and scheduled execution of discovery/evolution tasks without clear warnings, approval gates, or trust controls for external content. This is particularly dangerous because the skill’s purpose is to fetch, evaluate, and potentially install third-party skills, creating a supply-chain and persistence risk if unsafe content is installed or recurring tasks run without informed consent.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The activation description is broad enough to trigger on generic requests about skill management, upgrades, evaluation, or search, which can cause the skill to engage in situations beyond the user’s specific intent. In the context of a self-improvement and installation-oriented skill, overbroad activation raises the risk of unnecessary system modifications, unsafe recommendations, or unreviewed workflow takeover.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The module title and descriptive text are entirely in Chinese and present the skill identity and behavior in a fixed locale, with no indication that users can select another language. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale restriction is documented and justified.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The manifest describes a broad automation system for skill management, search, evaluation, installation, and daily evolution workflows. This file implements only simple local bookkeeping methods and a synthetic test suite, with no code for skill discovery, external search, evaluation engines, installation, or workflow orchestration.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The method name and returned field dependencies_checked imply a real integration/dependency validation step. In practice, it only returns a static dictionary and does not inspect or verify any dependencies, which makes the documented intent misleading relative to actual behavior.

Content

No source excerpt is available for this finding.