Security hardening and fixes are expected to focus on the active 1.x line once 1.0.0 is released. Pre-1.0 versions may still receive fixes, but that should not be treated as a long-term guarantee.

Please do not publish detailed exploit information in a public issue first.

Instead:

1. Prepare a concise report describing the issue, impact, and affected areas.
2. Include reproduction details only as needed for maintainers to verify it.
3. Share the report privately with the package maintainers through the repository’s preferred private disclosure channel.

If no dedicated security contact is configured yet, treat the issue as maintainer-private until a formal disclosure channel is published.

Security-sensitive areas for this package include:

• provider credential handling
• logging and redaction behavior
• cache isolation and cache key correctness
• multi-tenant safety assumptions in higher-level integrations
• tool execution and tool output handling
• replay/event persistence containing potentially sensitive data

• avoid introducing secret-bearing fixtures into committed test data
• prefer sanitized payload examples in docs and examples
• document any change that affects redaction, replay, persistence, or caching
• raise concerns early if a new feature weakens tenant or data isolation