Deterministic pre-release gate — four agents run real bandit, ruff, pytest, and docstring checks against hard thresholds

Notes & tips

• Depend on the documented public surface. The supported API is ReleasePrepTeamWorkflow, ReleasePrepTeam, and the ReleaseReadinessReport it returns. Names with a leading underscore — _evaluate_quality_gates, _run_command, _execute_tier — are internal.
• Gate after you draft. Use release-notes to draft the changelog, then release-prep to gate the ship on measured numbers.
• Keep it free. The default rule-based mode costs $0. Only set RELEASE_LLM_MODE=real when you want LLM-nuanced security/quality classification.
• Read the verdict, not the exit code. metadata["approved"] is the pass/fail; success only says the run completed.