Back to skill

Security audit

麻将陪练 · Mahjong Companion

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marked this skill suspicious because it found a browser rendering configuration that disables Chromium’s sandbox; the review is based on incomplete scanner coverage.

Review the sandbox-disabled Chromium finding before installing. The scan inspected 10 of 11 components; coverage was partial because the SKILL.md manifest could not be parsed, and static analysis did not cover the binary image. The results therefore do not establish that the skill is safe.

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
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
88% confidence
Finding

The subprocess call uses an argument list without a shell, so it does not itself introduce shell-command injection. However, the command explicitly disables Chromium's sandbox with --no-sandbox, removing an important containment boundary if rendered content exploits a browser vulnerability. Rendering locally generated Mahjong HTML reduces exposure, and this file alone does not establish an attacker-controlled HTML injection path.

Content

Scanner excerpt · scripts/render_hand.py (reported line 311)May include surrounding context.

python
f"--screenshot={out.resolve()}", f"--window-size={w},{h}",
        file_url,
    ]
    r = subprocess.run(cmd, capture_output=True, text=True, timeout=90)
    shutil.rmtree(tmp_profile, ignore_errors=True)

    if not out.exists():