test(ACTIVITY-WP-0016-T05): regression coverage incl. real 06-26 payload + over-depth

Add a test driving the actual captured 2026-06-26 failure payload
(tests/fixtures/wp0016/...partial.json): it now recovers 6+ valid recommendations
and quarantines the truncated tail, where before WP-0016 it discarded the whole run.
Add an over-depth guardrail test. Together with T03/T04 the regression set now covers
truncation, one-bad-item, oversized-string, over-depth, allow-list/injection-shaped,
and happy-path count cap.

In-repo portion of T05 complete; the live railiance01 graceful-degradation smoke is
operator-owned cluster work (deploy-coupled with the T02 bundle changes) and remains
outstanding. Hand-back notes posted to WP-0006-T03 and WP-0010-T04. Full suite: 220
passed, 1 skipped.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

tests/rules/test_executor.py

@@ -12,6 +12,7 @@ Covers:
 from __future__ import annotations
 
 import json
+from pathlib import Path
 from types import SimpleNamespace
 from typing import Any
 
@@ -531,6 +532,47 @@ def test_guardrail_allow_list_rejects_unknown_candidate():
     assert result.report["quarantined_items"][0]["reason"] == "allow_list"
 
 
+def _nested(depth: int) -> dict[str, Any]:
+    node: dict[str, Any] = {"leaf": 1}
+    for _ in range(depth):
+        node = {"a": node}
+    return node
+
+
+def test_guardrail_over_depth_quarantined():
+    deep = _valid_rec(2)
+    deep["extra"] = _nested(12)  # well past _MAX_DEPTH
+    raw = json.dumps({"summary": "Triage.", "recommendations": [_valid_rec(1), deep]})
+    llm = _CountingLLM([raw])
+
+    result = execute_instruction_with_audit(_triage_instr(), _Event(), {}, llm)
+
+    assert len(result.report["recommendations"]) == 1
+    assert result.report["quarantined_count"] == 1
+    assert result.report["quarantined_items"][0]["reason"] == "guardrail"
+    assert "depth" in result.report["quarantined_items"][0]["error"]
+
+
+def test_resilient_recovery_against_real_2026_06_26_fixture():
+    # The actual captured failure payload (4000-char preview, truncated at the 7th
+    # recommendation) — the run that reset the WP-0006-T03 streak. Before WP-0016
+    # this discarded the whole report; now it must recover the valid prefix.
+    fixture = json.loads(
+        Path("tests/fixtures/wp0016/daily_triage_2026-06-26_validation_failure.partial.json")
+        .read_text(encoding="utf-8")
+    )
+    raw = fixture["raw_output_preview"]
+    llm = _CountingLLM([raw, raw])
+
+    result = execute_instruction_with_audit(_triage_instr(), _Event(), {}, llm)
+
+    assert result.output_validated is True
+    assert result.report["partial"] is True
+    # Six recommendations are fully intact before the truncation point.
+    assert len(result.report["recommendations"]) >= 6
+    assert all("rank" in rec and "candidate" in rec for rec in result.report["recommendations"])
+
+
 def test_execute_instruction_with_audit_preserves_invalid_report_with_sinks(
     tmp_path,
     monkeypatch,

workplans/ACTIVITY-WP-0016-llm-output-robustness-trust-boundary.md

@@ -349,6 +349,25 @@ Done when:
   that the output-robustness blocker is cleared so the three-clean-run gate can
   resume on its own.
 
+2026-06-26 progress (in-repo portion complete):
+
+- **Regression coverage complete.** Across T03/T04/T05: truncated-mid-list,
+  one-bad-item-among-good (quarantine + partial), oversized-string and over-depth
+  guardrail rejection, allow-list (injection-shaped) rejection, happy-path count
+  cap, and a test driving the **actual captured 2026-06-26 payload**
+  (`tests/fixtures/wp0016/daily_triage_2026-06-26_validation_failure.partial.json`)
+  — it now recovers 6+ valid recommendations and quarantines the truncated tail,
+  where before it discarded the whole run.
+- **Full suite green:** 218 passed, 1 skipped (recorded at T04; the T05 fixture +
+  over-depth tests add to this — see the commit).
+- **Hand-back notes posted** to `ACTIVITY-WP-0006-T03` (State Hub event
+  `b6b8c2b8`) and `ACTIVITY-WP-0010-T04` (`b813f0dc`).
+- **Remaining (remote, operator-owned):** the live daily-triage smoke on
+  `railiance01` proving end-to-end graceful degradation. It depends on deploying
+  the T02 bundle prompt/`max_tokens`/NDJSON changes together with this code, which
+  is cluster/operator work outside this repo's SCOPE. T05 therefore stays
+  `progress` until that live run exists; the in-repo deliverables are done.
+
 ## Relationships
 
 - **Blocks / feeds:** `ACTIVITY-WP-0006-T03` (three clean scheduled runs) and