Skip to content

Security: TFD-42/Wild_Root_Prompt

SECURITY.md

Security Policy

Supported Versions

Version Supported
2.x (latest)
< 2.0

Reporting a Vulnerability

Please do not open a public issue for security vulnerabilities.

Use GitHub's private vulnerability reporting to submit a report confidentially. You will receive a response within 7 days.

Include:

  • Description of the vulnerability
  • Steps to reproduce
  • Potential impact
  • Suggested fix (optional)

Security Design

Wild Root Prompt is a local-first CLI tool. Key properties:

  • No telemetry — no data is sent to any remote server by default.
  • No credentials stored — API keys are read from environment variables or a local config file that is excluded from version control.
  • SSRF protection — the web server only accepts connections from localhost; all outbound fetch calls validate the target URL against a block-list before connecting.
  • Prompt injection detection — user input is scanned for known injection patterns before being forwarded to the model.
  • PII anonymization — the --anonymize flag strips emails, IPs, phone numbers, and other personal data from prompts before they leave the local machine.

See ZERO_TRUST_SECURITY.md for the full threat model.

There aren't any published security advisories