diff --git a/src/core/filter.rs b/src/core/filter.rs
index 1fc5a23f..2c223678 100644
--- a/src/core/filter.rs
+++ b/src/core/filter.rs
@@ -389,6 +389,27 @@ pub fn smart_truncate(content: &str, max_lines: usize, _lang: &Language) -> Stri
 mod tests {
     use super::*;
 
+    #[test]
+    fn test_filter() {
+        use crate::datatest;
+
+        datatest::run_tests("filter/*", |case| {
+            let level: FilterLevel = case
+                .params
+                .as_deref()
+                .expect("filter level required as params (no .after.<level>. in filename)")
+                .parse()
+                .expect("unknown filter level in filename");
+            let lang = Language::from_extension(
+                case.input_path
+                    .extension()
+                    .and_then(|e| e.to_str())
+                    .unwrap_or(""),
+            );
+            get_filter(level).filter(&case.input, &lang)
+        });
+    }
+
     #[test]
     fn test_filter_level_parsing() {
         assert_eq!(FilterLevel::from_str("none").unwrap(), FilterLevel::None);
@@ -480,20 +501,6 @@ mod tests {
         );
     }
 
-    #[test]
-    fn test_minimal_filter_removes_comments() {
-        let code = r#"
-// This is a comment
-fn main() {
-    println!("Hello");
-}
-"#;
-        let filter = MinimalFilter;
-        let result = filter.filter(code, &Language::Rust);
-        assert!(!result.contains("// This is a comment"));
-        assert!(result.contains("fn main()"));
-    }
-
     // --- truncation accuracy ---
 
     #[test]
diff --git a/src/datatest.rs b/src/datatest.rs
new file mode 100644
index 00000000..3b3e8d2b
--- /dev/null
+++ b/src/datatest.rs
@@ -0,0 +1,231 @@
+use std::collections::BTreeMap;
+use std::fs;
+use std::panic;
+use std::path::{Path, PathBuf};
+
+trait PathExt {
+    fn file_name_str(&self) -> &str;
+    fn file_stem_str(&self) -> &str;
+    fn extension_str(&self) -> &str;
+}
+
+impl PathExt for Path {
+    fn file_name_str(&self) -> &str {
+        self.file_name()
+            .expect("path has no file name")
+            .to_str()
+            .expect("non-UTF-8 path")
+    }
+    fn file_stem_str(&self) -> &str {
+        self.file_stem()
+            .expect("path has no stem")
+            .to_str()
+            .expect("non-UTF-8 path")
+    }
+    fn extension_str(&self) -> &str {
+        self.extension()
+            .expect("path has no extension")
+            .to_str()
+            .expect("non-UTF-8 path")
+    }
+}
+
+pub(crate) struct Case {
+    pub(crate) input: String,
+    pub(crate) expected: String,
+    /// The segment between `.after.` and the final extension.
+    /// `None` for files named `<name>.after.<ext>` with no params segment.
+    pub(crate) params: Option<String>,
+    pub(crate) input_path: PathBuf,
+    pub(crate) expected_path: PathBuf,
+}
+
+/// Runs test cases found under `tests/data/`.
+///
+/// **Path syntax**:
+/// - `"subdir/*"` — run all groups in `tests/data/subdir/`
+/// - `"subdir/prefix"` — run only groups whose name starts with `prefix`
+///   in `tests/data/subdir/` (group = prefix + extension, e.g. `language.rs`)
+///
+/// Each input file `<name>.<ext>` is paired with one or more expected files:
+///   `<name>.after.<ext>`               — no params
+///   `<name>.after.<params>.<ext>`      — with params
+///
+/// The closure returns the actual output; the engine compares it against the
+/// expected file (both sides trimmed).
+///
+/// **Update mode**: set `RTK_UPDATE_TEST_DATA=1` to overwrite expected files
+/// with the actual output instead of failing.  Re-run without the variable to
+/// confirm everything is green.
+///
+/// All failures are collected and reported together at the end.
+pub(crate) fn run_tests<F>(path: &str, test_fn: F)
+where
+    F: Fn(&Case) -> String + panic::RefUnwindSafe,
+{
+    let (dir_str, prefix_filter) = match path.strip_suffix("/*") {
+        Some(dir) => (dir, None),
+        None => match path.rfind('/') {
+            Some(i) => (&path[..i], Some(&path[i + 1..])),
+            None => (path, None),
+        },
+    };
+
+    let dir = Path::new(env!("CARGO_MANIFEST_DIR"))
+        .join("tests/data")
+        .join(dir_str);
+
+    let cases = discover(&dir, prefix_filter);
+    assert!(
+        !cases.is_empty(),
+        "no test cases found in {}{}",
+        dir.display(),
+        prefix_filter
+            .map(|p| format!(" (prefix: {p})"))
+            .unwrap_or_default()
+    );
+
+    let update = std::env::var("RTK_UPDATE_TEST_DATA").is_ok_and(|v| !v.is_empty());
+
+    let mut failures = Vec::new();
+    let mut updated = Vec::new();
+
+    for case in &cases {
+        let file_name = case.input_path.file_name_str();
+        let label = match &case.params {
+            Some(p) => format!("{file_name} [{p}]"),
+            None => file_name.to_string(),
+        };
+
+        let prev_hook = panic::take_hook();
+        panic::set_hook(Box::new(|_| {}));
+        let result = panic::catch_unwind(|| test_fn(case));
+        panic::set_hook(prev_hook);
+
+        match result {
+            Err(e) => {
+                let msg = e
+                    .downcast_ref::<String>()
+                    .map(String::as_str)
+                    .or_else(|| e.downcast_ref::<&str>().copied())
+                    .unwrap_or("unknown panic");
+                failures.push(format!("FAILED {label}:\n{msg}"));
+            }
+            Ok(actual) => {
+                if actual.trim() != case.expected.trim() {
+                    if update {
+                        let mut content = actual.trim().to_string();
+                        content.push('\n');
+                        fs::write(&case.expected_path, &content).unwrap_or_else(|e| {
+                            panic!("write {}: {}", case.expected_path.display(), e)
+                        });
+                        updated.push(label);
+                    } else {
+                        failures.push(format!(
+                            "FAILED {label}:\n--- expected\n{}\n+++ actual\n{}",
+                            case.expected.trim(),
+                            actual.trim()
+                        ));
+                    }
+                }
+            }
+        }
+    }
+
+    if !updated.is_empty() {
+        panic!(
+            "{} expected file(s) updated — re-run without RTK_UPDATE_TEST_DATA to verify:\n  {}",
+            updated.len(),
+            updated.join("\n  ")
+        );
+    }
+
+    assert!(
+        failures.is_empty(),
+        "{} case(s) failed:\n\n{}\n\n(set RTK_UPDATE_TEST_DATA=1 to overwrite)",
+        failures.len(),
+        failures.join("\n\n")
+    );
+}
+
+fn discover(dir: &Path, prefix_filter: Option<&str>) -> Vec<Case> {
+    // Group files by the name prefix before the first dot.
+    // With a prefix filter, only the matching group is processed;
+    // within that group each distinct extension forms its own input+expected set.
+    let mut groups: BTreeMap<String, Vec<PathBuf>> = BTreeMap::new();
+    for entry in fs::read_dir(dir).unwrap_or_else(|e| panic!("read dir {}: {}", dir.display(), e)) {
+        let path = entry
+            .unwrap_or_else(|e| panic!("read dir entry in {}: {}", dir.display(), e))
+            .path();
+        if !path.is_file() || path.extension().is_none() {
+            continue;
+        }
+        if let Some(name) = path.file_name().and_then(|n| n.to_str()) {
+            let prefix = name.split('.').next().unwrap_or(name).to_string();
+            if prefix_filter.is_none_or(|f| prefix == f) {
+                groups.entry(prefix).or_default().push(path);
+            }
+        }
+    }
+
+    let mut cases = Vec::new();
+
+    for (prefix, mut paths) in groups {
+        paths.sort();
+
+        let inputs: Vec<&PathBuf> = paths
+            .iter()
+            .filter(|p| {
+                !p.file_name()
+                    .and_then(|n| n.to_str())
+                    .map(|n| n.contains(".after."))
+                    .unwrap_or(false)
+            })
+            .collect();
+
+        if inputs.is_empty() {
+            panic!(
+                "no input file for group '{}' in {} (found only .after.* files)",
+                prefix,
+                dir.display()
+            );
+        }
+
+        for input_path in inputs {
+            let ext = input_path.extension_str().to_string();
+            let stem = input_path.file_stem_str().to_string();
+            let input = fs::read_to_string(input_path)
+                .unwrap_or_else(|e| panic!("read {}: {}", input_path.display(), e));
+
+            let after_prefix = format!("{}.after.", stem);
+            let ext_suffix = format!(".{}", ext);
+
+            for expected_path in paths.iter().filter(|p| {
+                p.file_name()
+                    .and_then(|n| n.to_str())
+                    .map(|n| n.starts_with(&after_prefix) && n.ends_with(&ext_suffix))
+                    .unwrap_or(false)
+            }) {
+                let expected_name = expected_path.file_name_str();
+                let middle = expected_name
+                    .strip_prefix(&after_prefix)
+                    .and_then(|s| s.strip_suffix(&ext_suffix))
+                    .unwrap_or("");
+                let params = (!middle.is_empty()).then(|| middle.to_string());
+
+                let expected = fs::read_to_string(expected_path)
+                    .unwrap_or_else(|e| panic!("read {}: {}", expected_path.display(), e));
+
+                cases.push(Case {
+                    input: input.clone(),
+                    expected,
+                    params,
+                    input_path: input_path.clone(),
+                    expected_path: expected_path.clone(),
+                });
+            }
+        }
+    }
+
+    cases
+}
diff --git a/src/main.rs b/src/main.rs
index e43f9267..438fe520 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -1,6 +1,8 @@
 mod analytics;
 mod cmds;
 mod core;
+#[cfg(test)]
+mod datatest;
 mod discover;
 mod hooks;
 mod learn;
diff --git a/tests/data/filter/python.after.aggressive.py b/tests/data/filter/python.after.aggressive.py
new file mode 100644
index 00000000..ac79c43c
--- /dev/null
+++ b/tests/data/filter/python.after.aggressive.py
@@ -0,0 +1,22 @@
+import os
+import unittest
+from pathlib import Path
+class Greeting:
+    def __init__(self, name):
+    // ... implementation
+class _Internal:
+    def __init__(self, value):
+    // ... implementation
+def greet(name):
+    // ... implementation
+def _helper(name):
+    // ... implementation
+def __helper(name):
+    // ... implementation
+def farewell(name):
+    // ... implementation
+def test_greet():
+    // ... implementation
+class TestGreeting(unittest.TestCase):
+    def test_greet(self):
+    // ... implementation
diff --git a/tests/data/filter/python.after.minimal.py b/tests/data/filter/python.after.minimal.py
new file mode 100644
index 00000000..7ce542e1
--- /dev/null
+++ b/tests/data/filter/python.after.minimal.py
@@ -0,0 +1,35 @@
+import os
+import unittest
+from pathlib import Path
+
+MAX = 100
+
+class Greeting:
+    def __init__(self, name):
+        self.name = name
+
+class _Internal:
+    def __init__(self, value):
+        self.value = value
+
+def greet(name):
+    """
+    Return a greeting.
+    """
+    return "Hello, " + name
+
+def _helper(name):
+    return name.upper()
+
+def __helper(name):
+    return name.lower()
+
+def farewell(name):
+    return "Goodbye, " + name
+
+def test_greet():
+    assert greet("World") == "Hello, World"
+
+class TestGreeting(unittest.TestCase):
+    def test_greet(self):
+        self.assertEqual(greet("World"), "Hello, World")
diff --git a/tests/data/filter/python.py b/tests/data/filter/python.py
new file mode 100644
index 00000000..5eac0033
--- /dev/null
+++ b/tests/data/filter/python.py
@@ -0,0 +1,37 @@
+# Module comment
+import os
+import unittest
+from pathlib import Path
+
+MAX = 100
+
+class Greeting:
+    def __init__(self, name):
+        self.name = name
+
+class _Internal:
+    def __init__(self, value):
+        self.value = value
+
+def greet(name):
+    """
+    Return a greeting.
+    """
+    # Inline comment.
+    return "Hello, " + name
+
+def _helper(name):
+    return name.upper()
+
+def __helper(name):
+    return name.lower()
+
+def farewell(name):
+    return "Goodbye, " + name
+
+def test_greet():
+    assert greet("World") == "Hello, World"
+
+class TestGreeting(unittest.TestCase):
+    def test_greet(self):
+        self.assertEqual(greet("World"), "Hello, World")
diff --git a/tests/data/filter/rust.after.aggressive.rs b/tests/data/filter/rust.after.aggressive.rs
new file mode 100644
index 00000000..5c70bf17
--- /dev/null
+++ b/tests/data/filter/rust.after.aggressive.rs
@@ -0,0 +1,15 @@
+use std::fmt;
+pub const MAX: usize = 100;
+pub struct Greeting {
+    // ... implementation
+pub fn greet(name: &str) -> String {
+    // ... implementation
+pub fn farewell(name: &str) -> String {
+    // ... implementation
+fn private_fn(name: &str) -> String {
+    // ... implementation
+fn test_farewell() {
+    // ... implementation
+    use super::*;
+    fn test_greet() {
+    // ... implementation
diff --git a/tests/data/filter/rust.after.minimal.rs b/tests/data/filter/rust.after.minimal.rs
new file mode 100644
index 00000000..47124c18
--- /dev/null
+++ b/tests/data/filter/rust.after.minimal.rs
@@ -0,0 +1,47 @@
+use std::fmt;
+
+/**
+ * Doc block kept by minimal.
+ */
+
+pub const MAX: usize = 100;
+
+pub struct Greeting {
+    name: String,
+}
+
+pub(crate) struct Internal {
+    value: u32,
+}
+
+/// Doc comment kept by minimal.
+pub fn greet(name: &str) -> String {
+    format!("Hello, {}", name)
+}
+
+pub(crate) fn helper(name: &str) -> String {
+    format!("Help, {}", name)
+}
+
+pub fn farewell(name: &str) -> String {
+    format!("Goodbye, {}", name)
+}
+
+fn private_fn(name: &str) -> String {
+    format!("Private, {}", name)
+}
+
+#[test]
+fn test_farewell() {
+    assert_eq!(farewell("World"), "Goodbye, World");
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_greet() {
+        assert_eq!(greet("World"), "Hello, World");
+    }
+}
diff --git a/tests/data/filter/rust.rs b/tests/data/filter/rust.rs
new file mode 100644
index 00000000..15dda63d
--- /dev/null
+++ b/tests/data/filter/rust.rs
@@ -0,0 +1,56 @@
+// Top-level comment
+use std::fmt;
+
+/* Block comment */
+
+/*
+ * Multi-line
+ * block comment
+ */
+
+/**
+ * Doc block kept by minimal.
+ */
+
+pub const MAX: usize = 100;
+
+pub struct Greeting {
+    name: String,
+}
+
+pub(crate) struct Internal {
+    value: u32,
+}
+
+/// Doc comment kept by minimal.
+pub fn greet(name: &str) -> String {
+    // Inline comment removed.
+    format!("Hello, {}", name)
+}
+
+pub(crate) fn helper(name: &str) -> String {
+    format!("Help, {}", name)
+}
+
+pub fn farewell(name: &str) -> String {
+    format!("Goodbye, {}", name)
+}
+
+fn private_fn(name: &str) -> String {
+    format!("Private, {}", name)
+}
+
+#[test]
+fn test_farewell() {
+    assert_eq!(farewell("World"), "Goodbye, World");
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_greet() {
+        assert_eq!(greet("World"), "Hello, World");
+    }
+}