Skip to content

Security: Advance-Labs/aeo-toolkit

Security

SECURITY.md

Security Policy

Reporting

Report vulnerabilities privately to the Advance Labs maintainers rather than opening a public issue.

Baseline guarantees

  • No secrets in git. .gitignore blocks .env*, *.pem, *.key, service-account*.json, credentials*.json. CI does not echo secrets.
  • BYOK keys are request-scoped. Users' LLM / Perplexity / OpenAI keys are never persisted or logged; they are read from the request and discarded.
  • OAuth tokens are encrypted at rest via a pluggable TokenStore interface (the production Supabase adapter encrypts refresh tokens; the default in-memory store is for local/dev only).
  • MCP servers enforce rate limiting (token bucket via @advance-labs/mcp-core) and return structured errors rather than leaking stack traces.
  • The crawler is polite: respects robots.txt, identifies itself via User-Agent, and rate-limits per host.
  • GitHub Actions never interpolate untrusted input into run: steps; secrets come from Actions secrets.

Dependency hygiene

  • Apache-2.0-licensed throughout; no AGPL/proprietary source was copied (see docs/adr/0002-typescript-from-scratch.md).
  • pnpm with a committed lockfile; CI installs with --frozen-lockfile.

There aren't any published security advisories