Skip to content

Security: MeterApp/skillhook

SECURITY.md

Security policy

skillhook exposes an HTTP endpoint that runs coding agents with real credentials on your machine, so we take reports seriously.

  • Report privately to hello@meterapp.co with steps to reproduce. Do not open a public issue for vulnerabilities. We aim to acknowledge within 3 business days.
  • Supported versions: the latest minor release on npm.

What skillhook does to protect you

  • Binds to 127.0.0.1; TLS and public exposure are delegated to Tailscale Funnel/Serve (or your own tunnel).
  • Every skill requires authentication by default (per-skill bearer token) and supports provider signatures (GitHub, Sentry, Linear, Standard Webhooks/Granola/Svix, Stripe, Slack) with timing-safe comparison and replay windows.
  • Secrets live in ~/.skillhook/.env (mode 600), are never logged, and are only handed to the agent when a skill lists them explicitly.
  • Webhook payloads are passed to the agent as clearly delimited untrusted data with guardrails; rate limits, body-size limits, per-skill filters and delivery de-duplication apply before anything runs.

See docs/security.md for the full model and recommendations.

There aren't any published security advisories