diff --git a/.kiro/specs/contributor-scan/tasks.md b/.kiro/specs/contributor-scan/tasks.md
index a470883..cf62689 100644
--- a/.kiro/specs/contributor-scan/tasks.md
+++ b/.kiro/specs/contributor-scan/tasks.md
@@ -197,6 +197,18 @@
   - _Requirements: 5.1, 8.1_
   - _Depends: 2.4_
 
+- [ ] 2.11 The suspicion band has to be worth reading
+  - libpng's scan produced 198 rows: 26 entailed (all one false positive, since fixed), 113 `c-unsafe-*`
+    pattern matches that assert an API is PRESENT rather than misused, 50 overlay-coverage annotations that
+    are not findings at all, and 9 model suspicions. Zero validated defects.
+  - The rung is honest -- these say "needs review" and mean it -- but a contributor handed 163 rows about a
+    library that uses `memcpy` correctly throughout will not read the one that matters. 82 of the 113 are in
+    `contrib/`, so task 2.10 removes most of it; what remains needs a policy, not a filter invented here.
+  - Observable: a stated rule for what reaches a contributor's report at `suspicion`, and libpng's row count
+    under it, measured.
+  - _Requirements: 5.1, 5.2_
+  - _Depends: 2.10_
+
 - [ ] 2.10 Do not spend the scan on example code
   - 22 of libpng's 30 highest-ranked regions are under `contrib/` -- the sample programs shipped with the
     library rather than the library. The budget and the 243s taint query go there first.
diff --git a/src/openultrasast/cpg/queries/taint.sc b/src/openultrasast/cpg/queries/taint.sc
index 800e1c3..0ab4af6 100644
--- a/src/openultrasast/cpg/queries/taint.sc
+++ b/src/openultrasast/cpg/queries/taint.sc
@@ -57,13 +57,16 @@
   // Word-boundary matching, never substring. `resolveUrl` contains `resolve`, so a bare-substring sanitizer
   // test marked the VULNERABLE flow sanitized and the verdict fell back to a captured `res` that names no
   // attacker input. Same bug class as the dominance query's guard matching; fixed in both.
+  // The boundary only applies where there IS a word character to bound: a token like `input(` ends in `(`,
+  // and an unconditional lookahead would make it unmatchable in `input(x)`.
+  def boundedPattern(token: String): java.util.regex.Pattern = {
+    val before = if (token.headOption.exists(c => c.isLetterOrDigit || c == '_')) "(?<![A-Za-z0-9_])" else ""
+    val after = if (token.lastOption.exists(c => c.isLetterOrDigit || c == '_')) "(?![A-Za-z0-9_])" else ""
+    java.util.regex.Pattern.compile(before + java.util.regex.Pattern.quote(token) + after)
+  }
+
   def mentionsToken(code: String, tokens: List[String]): Boolean =
-    tokens.exists(t =>
-      java.util.regex.Pattern
-        .compile("(?<![A-Za-z0-9_])" + java.util.regex.Pattern.quote(t) + "(?![A-Za-z0-9_])")
-        .matcher(code)
-        .find()
-    )
+    tokens.exists(t => boundedPattern(t).matcher(code).find())
 
   val sourcePatterns = split(sourcesS)
   val sinkNames      = split(sinksS)
@@ -145,10 +148,18 @@
 
   // A sink call: matched by short name (`system`), by leading code (`os.system(...)`), or by resolved
   // full name (`os.py:<module>.system`), so both bare and dotted forms in the spec hit.
+  //
+  // Every clause is word-BOUNDED, and the unbounded `contains` that used to end this list is why: `fgets`
+  // contains `gets`, so every bounded read in libpng matched the sink for the unbounded one, and all 26
+  // entailed findings on that repository were the same false positive -- `*argv -> fgets(buf, 256, f)`,
+  // which is the safe API being reported as the dangerous one. Third instance of this bug class here, after
+  // the sanitizer that matched `resolve` inside `resolveUrl` and the discharger that matched `user` inside
+  // `users`; the sink matcher had simply never been given the same treatment.
+  def sinkMatches(c: io.shiftleft.codepropertygraph.generated.nodes.Call, n: String): Boolean =
+    c.name == n || c.code.startsWith(n + "(") || mentionsToken(c.code, List(n)) || mentionsToken(c.methodFullName, List(n))
+
   def sinkCalls = {
-    val all = cpg.call.filter(c =>
-      sinkNames.exists(n => c.name == n || c.code.startsWith(n + "(") || c.code.startsWith(n) || c.methodFullName.contains(n))
-    )
+    val all = cpg.call.filter(c => sinkNames.exists(n => sinkMatches(c, n)))
     if (function.isEmpty) {
       // A region with no enclosing function: the file IS the scope.
       if (fileS.isEmpty) all else all.filter(_.method.filename.endsWith(fileS))
diff --git a/tests/test_model_specs.py b/tests/test_model_specs.py
index 0eb6330..a9676df 100644
--- a/tests/test_model_specs.py
+++ b/tests/test_model_specs.py
@@ -221,3 +221,23 @@ def test_a_guard_discharges_only_the_obligation_it_is_a_guard_of() -> None:
     assert "token_validator" not in identity, "authentication is not an object-level check"
     assert "token_validator" in set(spec.dischargers_by_kind["path_guard"])
     assert "['sub']" in identity or '["sub"]' in identity
+
+
+def test_a_sink_is_matched_as_a_word_not_a_substring() -> None:
+    """contributor-scan: `fgets` contains `gets`, and the safe API is not the dangerous one.
+
+    The sink matcher's last clause was an unbounded `methodFullName.contains(n)`, so every bounded read in
+    libpng matched the sink for the unbounded one: all 26 entailed findings on that repository were the same
+    false positive, `*argv -> fgets(buf, 256, f)`. Third instance of this bug class here, after the sanitizer
+    that matched `resolve` inside `resolveUrl` and the discharger that matched `user` inside `users`.
+
+    The query is Scala, so this asserts the fact tables still carry the pair that made it visible; the
+    boundary itself is exercised against a real CPG in the repository measurements.
+    """
+    from openultrasast.model.specs import taint_specs
+
+    memory = taint_specs(language="c")["memory"]
+
+    assert "gets" in memory.sinks, "the unbounded read is a sink"
+    assert "fgets" in memory.sources, "the bounded read is a SOURCE of untrusted input, never a sink"
+    assert "fgets" not in memory.sinks