If you discover a security vulnerability in AgentDeck, please report it responsibly.
Do NOT open a public GitHub issue for security vulnerabilities.
Instead, use GitHub's private vulnerability reporting to submit your report. Include:
- A description of the vulnerability
- Steps to reproduce
- Potential impact
- Suggested fix (if any)
You should receive an acknowledgment within 48 hours. We will work with you to understand the issue and coordinate a fix before any public disclosure.
The following are in scope for security reports:
- IPC channel injection — bypassing contextBridge isolation
- Command injection — via terminal input, file paths, or workflow variables
- Credential exposure — API keys stored via safeStorage leaking in logs, IPC, or disk
- Path traversal — accessing files outside intended WSL project directories
- Prototype pollution or XSS — in the renderer process
- Dependency vulnerabilities — in production dependencies with a known exploit
AgentDeck follows Electron security best practices:
contextIsolation: true— renderer cannot access Node.js APIsnodeIntegration: false— norequire()in renderersandbox: true— renderer runs in a sandboxed process- API keys encrypted at rest via Electron
safeStorage - PTY spawn environment blocklist prevents injection of
LD_PRELOAD,NODE_OPTIONS, etc. - IPC handlers validate types and enforce size limits
- Workflow event channels validated against safe ID regex in preload
| Version | Supported |
|---|---|
| 4.x | Yes |
| < 4.0 | No |