Open a GitHub issue with the security label: https://github.com/ktech7moon/agent-habitat/issues.

For private disclosure, reach out via the contact path on the maintainer's GitHub profile: https://github.com/ktech7moon.

This is a single-maintainer project. Public issues are the primary path; private email is available if the nature of the finding warrants it.

Pre-1.0. Only main is supported. Findings against older commits will be assessed against current main — no backports.

Issues that affect the framework's audit-chain integrity:

• Ways to bypass or subvert the critic's substring-grounding check (the fabrication contract, ADR-006 §3).
• Ways to corrupt the audit chain (workflows / workflow_steps / events rows, JSONL telemetry, or LangGraph checkpoint blobs) such that the trail no longer reflects what the framework actually did.
• Ways to cause the framework to log telemetry that should not be logged, or to leak telemetry across workflows.
• Dependency vulnerabilities reachable from default install (pip install -e ".[dev]").

Concerns documented as deployment responsibilities in the README's Production Considerations section:

• PII redaction at the observability layer (the framework ships the mechanism, not a redaction rule set).
• Reviewer authentication, identity, and notification fan-out for the human-in-the-loop checkpoint primitive.
• Operator-supplied configuration choices (rubric content, budget caps, API key handling outside the framework).

Best-effort, single-maintainer project. No formal SLA. Acknowledgement within a reasonable window; remediation timeline depends on severity and reproducibility.