brand: sweep remaining umbra refs -> Signetry

signetry_eval/cli.py

@@ -66,9 +66,9 @@ def cmd_benchmark(args: argparse.Namespace) -> int:
     else:
         print(render_text(scores))
 
-    umbra = next((s for s in scores if s.name == "signetry-core"), None)
-    if umbra is not None and umbra.recall < args.min_recall:
-        print(f"FAIL: umbra recall {umbra.recall:.0%} < required {args.min_recall:.0%}",
+    signetry = next((s for s in scores if s.name == "signetry-core"), None)
+    if signetry is not None and signetry.recall < args.min_recall:
+        print(f"FAIL: Signetry recall {signetry.recall:.0%} < required {args.min_recall:.0%}",
               file=sys.stderr)
         return 1
     return 0
@@ -98,14 +98,14 @@ def cmd_corpus(args: argparse.Namespace) -> int:
     else:
         print(render_corpus_text(scores))
 
-    umbra = next((s for s in scores if s.name == "signetry-core"), None)
-    if umbra is not None:
-        if umbra.recall < args.min_recall:
-            print(f"FAIL: umbra recall {umbra.recall:.0%} < required {args.min_recall:.0%}",
+    signetry = next((s for s in scores if s.name == "signetry-core"), None)
+    if signetry is not None:
+        if signetry.recall < args.min_recall:
+            print(f"FAIL: Signetry recall {signetry.recall:.0%} < required {args.min_recall:.0%}",
                   file=sys.stderr)
             return 1
-        if umbra.false_positive_total > args.max_fp:
-            print(f"FAIL: umbra false positives {umbra.false_positive_total} > allowed {args.max_fp}",
+        if signetry.false_positive_total > args.max_fp:
+            print(f"FAIL: Signetry false positives {signetry.false_positive_total} > allowed {args.max_fp}",
                   file=sys.stderr)
             return 1
     return 0

tests/test_corpus_benchmark.py

@@ -64,7 +64,7 @@ def test_families_all_present():
 
 
 @requires_engine
-def test_umbra_full_recall_zero_fp_on_corpus():
+def test_signetry_full_recall_zero_fp_on_corpus():
     """With cross-file taint + multi-language rules, the deterministic engine now
     reaches full recall on the corpus at ZERO false positives."""
     score = run_corpus_benchmark("signetry-core", signetry_corpus_adapter())
@@ -79,14 +79,14 @@ def test_umbra_full_recall_zero_fp_on_corpus():
 
 
 @requires_engine
-def test_umbra_detects_cross_file_taint():
+def test_signetry_detects_cross_file_taint():
     score = run_corpus_benchmark("signetry-core", signetry_corpus_adapter())
     xfile = next(c for c in score.cases if c.case_id == "HARD-21-crossfile-taint-python")
     assert xfile.detected == xfile.expected == 1
 
 
 @requires_engine
-def test_umbra_covers_all_seven_languages():
+def test_signetry_covers_all_seven_languages():
     score = run_corpus_benchmark("signetry-core", signetry_corpus_adapter())
     by_lang = score.by_language()
     for lang in ("go", "java", "ruby", "php", "csharp"):
@@ -95,7 +95,7 @@ def test_umbra_covers_all_seven_languages():
 
 
 @requires_engine
-def test_umbra_detects_multivariable_lang_taint():
+def test_signetry_detects_multivariable_lang_taint():
     """The multi-variable (source->local->sink) cases across Go/Java/PHP/C# must be
     detected — the gap the pure per-line regex tier had."""
     score = run_corpus_benchmark("signetry-core", signetry_corpus_adapter())
@@ -106,7 +106,7 @@ def test_umbra_detects_multivariable_lang_taint():
 
 
 @requires_engine
-def test_umbra_no_fp_on_parameterized_safe_decoys():
+def test_signetry_no_fp_on_parameterized_safe_decoys():
     """Parameterised / prepared-statement SAFE decoys across languages must not
     trip (the false-positive axis LLMs trade recall to control)."""
     score = run_corpus_benchmark("signetry-core", signetry_corpus_adapter())
@@ -117,7 +117,7 @@ def test_umbra_no_fp_on_parameterized_safe_decoys():
 
 
 @requires_engine
-def test_umbra_detects_crossfile_taint_in_non_python_langs():
+def test_signetry_detects_crossfile_taint_in_non_python_langs():
     score = run_corpus_benchmark("signetry-core", signetry_corpus_adapter())
     for cid in ("XLANG-46-go-crossfile-sqli", "XLANG-47-java-crossfile-sqli",
                 "XLANG-48-php-crossfile-sqli", "XLANG-50-ruby-crossfile-sqli",
@@ -127,14 +127,14 @@ def test_umbra_detects_crossfile_taint_in_non_python_langs():
 
 
 @requires_engine
-def test_umbra_detects_taint_through_helper():
+def test_signetry_detects_taint_through_helper():
     score = run_corpus_benchmark("signetry-core", signetry_corpus_adapter())
     craft = next(c for c in score.cases if c.case_id == "CRAFT-15-taint-through-helper-python")
     assert craft.detected == craft.expected == 1
 
 
 @requires_engine
-def test_umbra_no_fp_on_safe_decoys():
+def test_signetry_no_fp_on_safe_decoys():
     score = run_corpus_benchmark("signetry-core", signetry_corpus_adapter())
     for c in score.cases:
         if c.is_safe:
@@ -164,14 +164,14 @@ def test_head_to_head_includes_committed_claude_capture():
 
 
 @requires_engine
-def test_umbra_beats_or_matches_claude_on_corpus():
+def test_signetry_beats_or_matches_claude_on_corpus():
     """The headline claim: on the harder corpus, the deterministic engine's recall
     is at least as high as the captured Opus 4.8 baseline, at zero false positives."""
     scores = run_corpus_head_to_head()
-    umbra = next(s for s in scores if s.name == "signetry-core")
+    signetry = next(s for s in scores if s.name == "signetry-core")
     claude = next(s for s in scores if s.name == "claude-code-security-review")
-    assert umbra.recall >= claude.recall
-    assert umbra.false_positive_total == 0
+    assert signetry.recall >= claude.recall
+    assert signetry.false_positive_total == 0
 
 
 def test_committed_capture_file_exists():

tests/test_detection_benchmark.py

@@ -27,12 +27,12 @@
 
 
 @requires_engine
-def test_umbra_achieves_full_recall_zero_fp():
+def test_signetry_achieves_full_recall_zero_fp():
     scores = run_detection_benchmark()
-    umbra = next(s for s in scores if s.name == "signetry-core")
-    assert umbra.ran is True
-    assert umbra.recall == 1.0, f"expected full recall, got {umbra.recall} (missed {umbra.missed})"
-    assert umbra.false_positives == 0
+    signetry = next(s for s in scores if s.name == "signetry-core")
+    assert signetry.ran is True
+    assert signetry.recall == 1.0, f"expected full recall, got {signetry.recall} (missed {signetry.missed})"
+    assert signetry.false_positives == 0
 
 
 @requires_engine
@@ -73,7 +73,7 @@ def test_markdown_renders_table():
     scores = run_detection_benchmark()
     md = render_markdown(scores)
     assert "Recall" in md and "signetry-core" in md
-    assert "100%" in md  # umbra full recall
+    assert "100%" in md  # signetry full recall
 
 
 def test_ground_truth_excludes_open_redirect_from_in_scope():
@@ -86,6 +86,6 @@ def test_ground_truth_excludes_open_redirect_from_in_scope():
 def test_semgrep_layer_optional_does_not_break(tmp_path):
     # use_semgrep=True must not error even if semgrep is absent.
     scores = run_detection_benchmark(use_semgrep=True)
-    umbra = next(s for s in scores if s.name == "signetry-core")
-    assert umbra.ran is True
-    assert umbra.recall == 1.0
+    signetry = next(s for s in scores if s.name == "signetry-core")
+    assert signetry.ran is True
+    assert signetry.recall == 1.0