Thanks for taking the time to look. Agency Agents writes files into developer-tool configuration directories, so security reports are welcome.

This is a pre-1.0 project. Support windows may change as release packaging stabilizes.

Email msitarzewski@gmail.com with:

• a clear description of the issue
• impact
• steps to reproduce or proof of concept
• version or commit tested
• name/handle for credit, if desired

Please do not open a public GitHub issue for security reports.

Best effort:

• acknowledgement within 7 days
• initial assessment within 14 days
• fix or mitigation plan within 30 days for high/critical findings

In scope:

• remote code execution in the app or IPC commands
• arbitrary file read/write through Tauri commands
• path traversal in install, backup, catalog, or import/export paths
• unsafe overwrite or uninstall behavior
• bypass of modified-file backup protections
• XSS in the webview
• token leakage from GitHub OAuth storage or IPC
• SSRF or unexpected outbound requests
• updater signature or artifact verification bypass
• incorrect tool path handling that writes outside documented destinations

Out of scope:

• vulnerabilities in third-party AI coding tools
• malicious agent content in a user-selected catalog clone
• vulnerabilities in macOS, Windows, Linux, WebKit, or system components
• social-engineering attacks
• attacks requiring an already-compromised local account
• bugs in the upstream agency-agents repo that do not affect app install/write behavior

Coordinated disclosure, 90-day default. If a fix takes longer, the reporter and maintainer can agree on an extension before the embargo expires.

• Typed Tauri IPC.
• No frontend arbitrary shell bridge.
• Deterministic renderers for supported tools.
• Local ledger records app-managed writes.
• Modified installed files are backed up before destructive operations.
• GitHub token is stored in the platform keychain and is not returned to the frontend.
• Network features are gated by settings and feature boundaries.

Reporters who have responsibly disclosed security issues:

(empty - be the first)