Skip to content

Security: codesark/ai-platform-lab

Security

SECURITY.md

Security Policy

This is a public, MIT-licensed reference project. It is built with public-repo safety as a first-class concern.

Reporting a vulnerability

Please report security issues privately via GitHub: Security → Report a vulnerability (private advisory) on this repository. Do not open a public issue for a suspected vulnerability.

How secrets are handled

  • No secret is ever committed. .env is gitignored; only .env.example (placeholders) is tracked.
  • gitleaks runs as a pre-commit hook and in CI, and fails closed.
  • Cloud credentials use OIDC / Workload Identity, never long-lived JSON keys committed to the repo.
  • Terraform state is stored in a remote, encrypted, locked backend and is never committed (*.tfstate is gitignored).
  • API keys (Gemini, Langfuse) live only in local .env and in GitHub Actions secrets.

Application security posture

  • Input validation + size limits on every endpoint (Pydantic).
  • API-key auth, rate-limiting, and CORS (locked to the UI origin) as FastAPI middleware.
  • The seed corpus is synthetic and public-safe — no PII.
  • Agent tools (planned) are allow-listed and argument-validated server-side; the MCP server authenticates its transport and treats tool descriptions/results as untrusted (tool-poisoning / prompt-injection defense).

There aren't any published security advisories