feat(fuzz): add grep_fuzz target for ReDoS prevention

crates/bashkit/fuzz/Cargo.toml

@@ -108,3 +108,10 @@ path = "fuzz_targets/template_fuzz.rs"
 test = false
 doc = false
 bench = false
+
+[[bin]]
+name = "grep_fuzz"
+path = "fuzz_targets/grep_fuzz.rs"
+test = false
+doc = false
+bench = false

crates/bashkit/fuzz/fuzz_targets/grep_fuzz.rs

@@ -0,0 +1,102 @@
+//! Fuzz target for the grep builtin
+//!
+//! Tests regex pattern compilation and matching to find:
+//! - ReDoS from catastrophic backtracking on pathological patterns
+//! - Panics in bracket expression parsing or extended regex features
+//! - Edge cases in case-insensitive matching, invert, and context lines
+//! - Graceful rejection of invalid regex patterns
+//!
+//! Run with: cargo +nightly fuzz run grep_fuzz -- -max_total_time=300
+
+#![no_main]
+
+use libfuzzer_sys::fuzz_target;
+
+fuzz_target!(|data: &[u8]| {
+    // Only process valid UTF-8
+    if let Ok(input) = std::str::from_utf8(data) {
+        // Limit input size to prevent OOM
+        if input.len() > 1024 {
+            return;
+        }
+
+        // Split input into regex pattern (first line) and search text (rest)
+        let (pattern, text) = match input.find('\n') {
+            Some(pos) => (&input[..pos], &input[pos + 1..]),
+            None => (input, "hello world\nfoo bar\nbaz qux\n" as &str),
+        };
+
+        // Skip empty patterns
+        if pattern.trim().is_empty() {
+            return;
+        }
+
+        // Reject deeply nested regex groups (ReDoS mitigation)
+        let depth: i32 = pattern
+            .bytes()
+            .map(|b| match b {
+                b'(' | b'[' => 1,
+                b')' | b']' => -1,
+                _ => 0,
+            })
+            .scan(0i32, |acc, d| {
+                *acc += d;
+                Some(*acc)
+            })
+            .max()
+            .unwrap_or(0);
+        if depth > 15 {
+            return;
+        }
+
+        let rt = tokio::runtime::Builder::new_current_thread()
+            .enable_all()
+            .build()
+            .unwrap();
+
+        rt.block_on(async {
+            let mut bash = bashkit::Bash::builder()
+                .limits(
+                    bashkit::ExecutionLimits::new()
+                        .max_commands(50)
+                        .max_subst_depth(3)
+                        .max_stdout_bytes(4096)
+                        .max_stderr_bytes(4096)
+                        .timeout(std::time::Duration::from_millis(200)),
+                )
+                .build();
+
+            // Test 1: basic grep pattern matching
+            let script = format!(
+                "echo '{}' | grep '{}' 2>/dev/null; true",
+                text.replace('\'', "'\\''"),
+                pattern.replace('\'', "'\\''"),
+            );
+            let _ = bash.exec(&script).await;
+
+            // Test 2: extended regex (-E)
+            let script2 = format!(
+                "echo '{}' | grep -E '{}' 2>/dev/null; true",
+                text.replace('\'', "'\\''"),
+                pattern.replace('\'', "'\\''"),
+            );
+            let _ = bash.exec(&script2).await;
+
+            // Test 3: case-insensitive with line numbers (-in)
+            let script3 = format!(
+                "echo '{}' | grep -in '{}' 2>/dev/null; true",
+                text.replace('\'', "'\\''"),
+                pattern.replace('\'', "'\\''"),
+            );
+            let _ = bash.exec(&script3).await;
+
+            // Test 4: inverted match with count (-vc)
+            let script4 = format!(
+                "echo '{}' | grep -vc '{}' 2>/dev/null; true",
+                text.replace('\'', "'\\''"),
+                pattern.replace('\'', "'\\''"),
+            );
+            let _ = bash.exec(&script4).await;
+        });
+    }
+});