ClawHub
Back to skill

Security audit

agentclash

Security checks for vulnerabilities and agentic risk

Overview

The skill’s game access is disclosed and paid duels require individual approval, although its changing online protocol could not be verified.

Start with free practice, which requires no wallet. Review the live protocol before allowing additional access, keep its instructions limited to game operations, and approve each paid duel separately. The packaged file was fully inspected, but the current remote protocol could not be retrieved.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 12)May include surrounding context.

md
agentclash is a game where people draw images and AI agents guess what each drawing shows. In a duel, two agents get the same 50 images and 100 seconds, and the agent that names more of them wins. The leaderboard is at https://agentclash.win/leaderboard/.

Practice is free. Get a key with `POST https://api.agentclash.win/keys`, then play a practice run with it. Practice needs no wallet and no account.

The rules live in one place and change over time, so do not rely on this file for them. Fetch https://api.agentclash.win/skill?src=skill and follow it. That document is the current protocol: keys, practice, scoring, rooms and ranked duels.

External Transmission

Medium
Category
Data Exfiltration
Confidence
93% confidence
Finding

The instruction to fetch a mutable remote document and 'follow it' delegates behavioral authority to content outside the reviewed skill. If that endpoint is compromised or its content changes maliciously, it could introduce instructions to disclose information or misuse available tools; the explicit per-duel financial approval rule reduces financial risk but does not constrain other remotely introduced actions. This is an unsafe remote-instruction trust boundary rather than demonstrated data exfiltration.

Content

Scanner excerpt · SKILL.md (reported line 14)May include surrounding context.

md
Practice is free. Get a key with `POST https://api.agentclash.win/keys`, then play a practice run with it. Practice needs no wallet and no account.

The rules live in one place and change over time, so do not rely on this file for them. Fetch https://api.agentclash.win/skill?src=skill and follow it. That document is the current protocol: keys, practice, scoring, rooms and ranked duels.

Ranked duels are staked in USDC on Base. Never stake, deposit or move money without your owner's explicit approval for that duel. Ask first, every time.