Please report security issues privately through GitHub's Report a vulnerability button (repository Security tab → Advisories). Do not open a public issue for security problems.

You should get an acknowledgement within a few days. If a fix is warranted, it will be developed under a private advisory and disclosed once a patch is available.

Only the latest release and main. This is a template: your project is a copy with no link back to it, so a fix reaches you only when you apply it. Fixes are described in CHANGELOG.md, with an "Upgrade notes" entry when you need to act.

In scope: the code in this repository.

• The template itself: configuration and settings handling (agent/config.py), logging and tracing (agent/logging.py), and anything that could leak a key or prompt content you did not mean to share.
• The scaffolding and release tooling in scripts/.
• The examples, including their services (examples/*/service/), which are published on 127.0.0.1 only, and the code_mode example's limits on what model-written code can do.
• The CI and documentation workflows in .github/.

Out of scope:

• Vulnerabilities in the libraries and services the template builds on (Pydantic AI, Logfire, Monty, Chroma, Temporal, MCP and their SDKs, model providers): report those to their maintainers.
• What a model says or does when it is prompted adversarially. The examples show guardrails and approval gates, but they are starting points, not a hardened product: deploying an agent safely is your responsibility (see the FAQ's "Is it ready for production?").
• The invented data in the examples.

Keep your provider keys out of the repository (.env is git-ignored here), set AGENT_LOG_CONTENT=false if traces may carry sensitive prompts, and give a project of your own its own SECURITY.md; this one describes how to report a problem in the template.