Skip to content

Security: emisso-ai/emisso-self-healing

Security

SECURITY.md

Security Policy

Reporting a Vulnerability

Do not open a public issue. Email hello@emisso.ai with:

  • Description of the vulnerability
  • Steps to reproduce
  • Potential impact

We will acknowledge within 48 hours and aim to fix critical issues within 7 days.

Sensitive Areas

This SDK creates GitHub PRs, executes code in sandboxes, and handles API tokens for multiple services. Issues in these areas are treated with highest priority:

  • Code execution — the sandbox runner executes AI-generated code (src/sandbox/)
  • GitHub access — PR creation with write permissions (src/repair/github.ts)
  • Token handling — Anthropic, GitHub, Vercel, Supabase, Slack tokens flow through the pipeline
  • Safety guardrails — any bypass of rate limits, confidence thresholds, or path exclusions (safety config)

Safety Invariants

These must never be weakened without explicit user opt-in:

  • Fixes are never auto-merged by default
  • Migrations, secrets, and .env files are always excluded
  • PRs are only created after sandbox tests pass
  • Rate limiting caps PRs per hour

Supported Versions

Version Supported
Latest Yes

There aren't any published security advisories