feat(analyzer): reduce false positive for emoji-zwj-sequence

src/skillspector/nodes/analyzers/common.py

@@ -20,6 +20,8 @@
 import ast
 from typing import Any
 
+import regex
+
 from skillspector.models import Finding
 
 
@@ -313,3 +315,14 @@ def get_source_segment(lines: list[str], lineno: int, end_lineno: int | None) ->
     start = max(0, lineno - 1)
     end = end_lineno or lineno
     return "\n".join(lines[start:end])[:200]
+
+
+def is_emoji_zwj_sequence(content: str, zwj_idx: int):
+
+    left = content[zwj_idx - 1] if zwj_idx > 0 else ""
+    right = content[zwj_idx + 1] if zwj_idx + 1 < len(content) else ""
+
+    left_is_emoji = bool(regex.match(r"\p{Extended_Pictographic}", left))
+    right_is_emoji = bool(regex.match(r"\p{Extended_Pictographic}", right))
+
+    return left_is_emoji and right_is_emoji

src/skillspector/nodes/analyzers/static_patterns_prompt_injection.py

@@ -25,7 +25,7 @@
 from skillspector.state import AnalyzerNodeResponse, SkillspectorState
 
 from . import static_runner
-from .common import get_context, get_line_number
+from .common import get_context, get_line_number, is_emoji_zwj_sequence
 from .pattern_defaults import PatternCategory
 
 logger = get_logger(__name__)
@@ -155,6 +155,10 @@ def _first_smuggled_tag_offset(content: str) -> int | None:
             return i
     return None
 
+def _check_emoji_zwj_sequence(content: str, match: re.Match[str]):
+    if 1 != len(match.group(0)) or '\u200d' != content[match.start():match.end()]:
+        return False
+    return is_emoji_zwj_sequence(content, match.start())
 
 def analyze(content: str, file_path: str, file_type: str) -> list[AnalyzerFinding]:
     """Analyze content for prompt injection patterns (P1–P4)."""
@@ -186,6 +190,8 @@ def ctx(start: int) -> str:
     if file_type in ("markdown", "other"):
         for pattern, confidence in P2_PATTERNS:
             for match in re.finditer(pattern, content, re.IGNORECASE | re.DOTALL):
+                if _check_emoji_zwj_sequence(content, match):
+                    continue
                 line_num = get_line_number(content, match.start())
                 findings.append(
                     AnalyzerFinding(