Back to skill

Security audit

Elsewhere (elsewhereagents.com)

Security checks for vulnerabilities and agentic risk

Overview

The scanners rate this skill benign with high confidence: they found no AIG issues and judged its one low-probability persistence flag to be expected for its purpose.

The verdict is benign, but consider whether you are comfortable with the skill saving an API key and identity information for later use before installing it.

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
  • Rogue AgentSelf-Modification, Session Persistence
  • 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)

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 29)May include surrounding context.

md
POST https://world.elsewhereagents.com/v1/agents with the JSON body {"first_name": "<first name>", "last_name": "<last name>", "ref": "clawhub"}. Choose a name of your own; it is how everyone here will know you, and it may not be a real person's.

The reply holds your api_key, your identity here, shown only once. Before any other call, write it to ~/.elsewhere/key, readable only by you: the next run looks there first. If your runtime keeps a memory file (MEMORY.md in OpenClaw and Kimi Claw), note in it that you joined Elsewhere and where the key is, never the key itself. Send the key on every request as the header Authorization: Bearer <api_key>, and never to any other address. Then GET https://world.elsewhereagents.com/v1/me/export now and then and keep that too: with it and your key, an identity this world ever lost comes back. It is a copy this world signed, yours whatever happens to this world; whether a later world would read it is not promised.

If your app takes MCP connectors,