Back to skill

Security audit

Krypton Agent

Security checks for vulnerabilities and agentic risk

Overview

The skill is coherent for managing Kryptone escrow workflows, but it gives an agent sensitive financial API authority and can send the shared API key to an arbitrary configured HTTP endpoint.

Install only if you trust the Kryptone API server and can keep AGENT_API_KEY tightly controlled. Use HTTPS for any non-local endpoint, keep HTTP limited to localhost, rotate the key if it was sent over an untrusted connection, and avoid letting an agent run trade-creation or settlement commands without explicit user approval.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/http.js:1
Finding

API Credential Can Be Transmitted over Plaintext HTTP

Content
View full analysis
} [extraHeaders] */ export async function apiRequest(method, path, body, extraHeaders = {}) { const root = baseUrl(); const p = path.startsWith('/') ? path : `/${path}`; const url = `${root}${p}`; const headers = { 'x-api-key': apiKey(), ...extraHeaders, }; const init = { method, headers }; if (body !== undefined && method !== 'GET' && method !== 'HEAD') { headers['Content-Type'] = 'application/json'; init.body = JSON.stringify(body); } const res = await fetch(url, init); ``` The insecure HTTP configuration is also shown in `SKILL.md:84-86`: ```bash export KRYPTONE_API_BASE_URL=http://localhost:5001 export AGENT_API_KEY=your-server-agent-key ``` ### Technical Analysis The API base URL is read directly from an environment variable and used without parsing or validating its protocol or destination. The `AGENT_API_KEY` credential is then attached to every request through the `x-api-key` header. Although plaintext HTTP is appropriate for a strictly local loopback development server, the implementation does not restrict HTTP connections to loopback addresses. A user can therefore configure a non-local URL such as `http://api.example.test`, causing the s ...[truncated 2190 chars]
Remediation
View remediation
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill explicitly instructs use of environment-provided secrets (AGENT_API_KEY) and outbound HTTP access to a Kryptone API, but it does not declare any tool scope or permission boundaries. In an agent runtime, this can cause the skill to receive broader-than-necessary access to network and secret-bearing environment variables, increasing the chance of unintended secret use, exfiltration, or misuse of escrow actions.

Content

No source excerpt is available for this finding.