Security Policy

Supported Versions

We provide security updates for the latest main branch and the latest tagged release.

Version	Supported
Latest release (v*)	✅
main branch	✅
Older releases	❌
Development branches/forks	❌

Reporting a Vulnerability

If you discover a security vulnerability in transcriptd, please report it privately. Do not create a public GitHub Issue.

To report a vulnerability, please use GitHub Security Advisories:

1. Go to the Security Advisories tab in this repository.
2. Click Report a Vulnerability.
3. Provide as much detail as possible.

Alternatively, you can email [email protected] directly.

What to include in your report

Please include:

• A description of the vulnerability and its impact
• Product version, OS, compiler version (if applicable)
• Steps to reproduce the issue, or a proof-of-concept
• Suggested remediation (optional but appreciated)

Response timeline

• Acknowledgment: We aim to acknowledge receipt of your vulnerability report within 48 hours.
• Triage: We aim to complete initial triage within 7 days.
• Disclosure: We will coordinate with you on public disclosure after a fix has been rolled out.

Scope

In Scope:

• The core transcriptd daemon, CLI, and TUI implementation in this repository
• The MCP server implementation (src/mcp.rs)
• The extraction adapters (crates/transcriptd-*)
• Inter-process communication logic (src/ipc.rs)
• Vulnerabilities in the shipped GitHub Actions CI pipelines (e.g. supply chain attacks)

Out of Scope:

• Vulnerabilities in the underlying IDEs (Zed, VS Code, Claude Code)
• Issues arising from local machine compromise (e.g. if root access was gained independently, the attacker can already read the AI databases directly)
• Features marked explicitly as experimental or prototype without production intent
• Unofficial forks or third-party builds

Common Weaknesses

When investigating transcriptd, please be mindful of these areas:

• Paths and symlinks: Extractors must not follow malicious symlinks or allow path traversal outside the user's workspace.
• Daemon exposure: The MCP HTTP endpoint and Unix domain socket must not inadvertently expose sensitive transcript data across OS user boundaries.
• XSS in TUI: Though terminal-based, maliciously crafted input from AI platforms should not lead to terminal escape sequence execution.
• SQL Injection: Though we use parameter binding in transcriptd-store, LIKE / GLOB bounds in FTS queries should remain strictly sanitized.