fix(confidence): pass multi_lens_observed kwarg to merge_confidence

pilot/ai_review.py

@@ -234,7 +234,7 @@ def _int_env(name: str, default: int) -> int:
 _CONFIDENCE_BADGE = {5: "🟢", 4: "🟢", 3: "🟡", 2: "🟠", 1: "🔴"}
 
 
-def merge_confidence(findings: list[dict]) -> int:
+def merge_confidence(findings: list[dict], *, multi_lens_observed: bool = False) -> int:
     """1-5 merge verdict: higher = safer.
 
     Tier drops driven by the most severe finding present:
@@ -243,9 +243,14 @@ def merge_confidence(findings: list[dict]) -> int:
       - medium    → 4
       - else      → 5 (low / trivial / info / unknown → no drop)
 
-    An extra -1 when ANY finding carries ``_multi_lens`` (cross-lens agreement
-    bumps confidence in the signal — penalise harder). The final score is
-    clamped to [1, 5] so a critical + multi_lens combo doesn't go negative.
+    An extra -1 when cross-lens agreement was observed on any finding
+    (``multi_lens_observed``). The flag is passed in explicitly because the
+    raw ``_multi_lens`` marker is stripped from findings by the time they
+    reach this function — first by ``opencode_review.run_lenses_review``
+    (the ``_``-prefix scrub) and again by ``_normalize_finding`` (the
+    7-key schema rebuild). The caller (``review_pr``) must capture the
+    signal before those strips fire. Final score is clamped to [1, 5] so
+    a critical + multi_lens combo doesn't go negative.
     """
     if not findings:
         return 5
@@ -258,7 +263,7 @@ def merge_confidence(findings: list[dict]) -> int:
         score = 4
     else:
         score = 5
-    if any(f.get("_multi_lens") for f in findings):
+    if multi_lens_observed:
         score -= 1
     return max(1, min(5, score))
 
@@ -2065,6 +2070,15 @@ def review_pr(
             })
         except Exception:
             changed_paths = []
+        # Capture cross-lens agreement BEFORE apply_repo_config — by the time
+        # findings land in `review_pr` the `_multi_lens` marker has already
+        # been scrubbed (once by `opencode_review.run_lenses_review`'s
+        # `_`-prefix strip, again by `_normalize_finding`'s 7-key rebuild),
+        # so `merge_confidence` cannot read it off the dict. We scan here as
+        # the convergence point for both engine paths; in practice the kwarg
+        # currently always passes False, but the structural plumbing is
+        # correct for any future code path that preserves the flag.
+        multi_lens = any(f.get("_multi_lens") for f in findings)
         kept, _dropped = apply_repo_config(findings, config, changed_paths=changed_paths)
         findings = kept
         if _dropped:
@@ -2111,7 +2125,10 @@ def review_pr(
         # anchoring so the verdict reflects what the operator sees (a critical
         # finding that fails to anchor is still a critical finding). The
         # default 5 keeps any failure path (e.g. empty findings) green.
-        confidence = merge_confidence(findings)
+        # Cross-lens agreement is passed in via kwarg (see multi_lens scan
+        # above) because the `_multi_lens` flag is stripped before findings
+        # reach this call.
+        confidence = merge_confidence(findings, multi_lens_observed=multi_lens)
         summary_body = format_review_body(
             "\n\n".join(summary_parts), model, sha,
             summary=review_summary,

tests/pilot/test_ai_review.py

@@ -1855,8 +1855,25 @@ def test_merge_confidence_critical_drops_to_one():
 
 
 def test_merge_confidence_multi_lens_drops_extra():
-    f = {"severity": "low", "_multi_lens": True}
-    assert merge_confidence([f]) == 4
+    # The flag has moved to a kwarg; passing `_multi_lens` on the dict is no
+    # longer enough — the kwarg is the only path that drops the score.
+    f = {"severity": "low"}
+    assert merge_confidence([f], multi_lens_observed=True) == 4
+
+
+def test_merge_confidence_multi_lens_survives_normalization():
+    """Real flow: `_multi_lens` is set on the raw finding, but stripped by
+    `_normalize_finding`. `merge_confidence(...)` with only the kwarg sees a
+    normalized finding; the dedup must be triggered by `multi_lens_observed=`
+    being true, not by reading `_multi_lens` off the dict."""
+    raw = {"_multi_lens": True, "severity": "low", "path": "x", "line": 1,
+           "problem": "p", "fix": "", "suggestion": "", "reference": ""}
+    normalized = _normalize_finding(raw)
+    assert "_multi_lens" not in normalized  # confirms the strip
+    # Now call merge_confidence the way review_pr will:
+    assert merge_confidence([normalized], multi_lens_observed=True) == 4
+    # And without the kwarg, the flag-on-dict path is gone:
+    assert merge_confidence([normalized]) == 5
 
 
 def test_merge_confidence_clamped():