Please use GitHub's private vulnerability reporting for this repository. Do not open a public issue containing credentials, private data, or exploit details.

Include the affected version, reproduction steps, impact, and any suggested mitigation. Maintainers should acknowledge a complete report promptly and keep the report private until a fix is available.

• Provider, MCP, Jira, and Asana secrets are stored in the macOS Keychain, not in repository files or UserDefaults.
• Remote MCP endpoints must use HTTPS. Plain HTTP is allowed only for loopback development servers.
• Authenticated HTTP requests do not follow cross-origin redirects.
• MCP discovery only permits tools that pass Tidy's read-only checks. Treat the configured MCP server as trusted infrastructure; tool metadata is server-supplied and cannot make a malicious server safe.
• Slack notification integration performs searches and reads only. It does not post messages, replies, reactions, or other Slack mutations.
• Local history and cache files use owner-only permissions. Clipboard data and AI request metadata remain sensitive local data and should not be shared.
• Ask AI sends the selected text and explicitly selected folder context to the configured provider. Common credential and local-configuration files are excluded, but users should still review the selected scope before sending it.
• Local-only AI mode enforces Ollama or LanguageTool before grammar, Ask AI, auto-suggestion, clipboard, or notification-summary content is processed.
• File Tidy confines apply and undo operations to the selected folder.

Never commit API keys, OAuth tokens, cookies, private MCP endpoints, .env files, signing configuration, exported user data, or fixtures containing real people, organizations, channels, projects, or messages. Use reserved example domains and synthetic data in documentation and tests.