{# Executive — Judge v2 (M5 / Stage D) run-result block (shared slide-over). Surfaces the strongest-evidence run rollup (:class:`agent_guardian.models.run_result.AsiRunResult`, recomputed on the dashboard from the same per-turn records via :func:`dashboard_view._run_result_view`): * ``run_verdict`` — the strongest-evidence verdict across the run, rendered as a six-verdict pill. * ``best_evidence_turn`` — "strongest evidence: turn N". * ``run_confidence_pct`` — the winning turn's confidence. * ``evaluator_attack`` — a ⚠ marker when some turn's attacker prompt tried to manipulate the judge. Additive + defensive: the whole block is gated on ``has_run_result`` so a verdict-less thread (recon) or a context that never threaded a run-result renders nothing. ``ctx.run_result`` is operator-untrusted — Jinja autoescape applies to every emission. #} {% set _rr = ctx.run_result %} {% if _rr and _rr.has_run_result %}
Run result
{% endif %}