Cloudflare triage inspired improvements

crates/claudear-analysis/src/evaluation/types.rs

@@ -29,6 +29,20 @@ impl EvaluationResult {
         }
     }
 
+    /// Whether any test tool gained new failures vs the baseline.
+    pub fn has_new_test_failures(&self) -> bool {
+        self.deltas
+            .iter()
+            .any(|d| d.after.category == EvalCategory::Test && d.new_failures > 0)
+    }
+
+    /// Whether the fix introduced new failures or regressions in any tool.
+    pub fn has_regressions(&self) -> bool {
+        self.deltas
+            .iter()
+            .any(|d| d.new_failures > 0 || !d.regressions.is_empty())
+    }
+
     fn build_summary(deltas: &[EvalDelta]) -> String {
         if deltas.is_empty() {
             return "No evaluation tools ran.".to_string();
@@ -225,6 +239,42 @@ mod tests {
         assert!(!result.summary.is_empty());
     }
 
+    #[test]
+    fn test_has_new_test_failures() {
+        // A newly-added failing test (red) shows up as a new test failure.
+        let red = EvaluationResult::new(
+            1,
+            "org/repo".into(),
+            vec![EvalDelta::compute(
+                make_snapshot(EvalCategory::Test, "cargo test", 10, 0),
+                make_snapshot(EvalCategory::Test, "cargo test", 10, 1),
+            )],
+        );
+        assert!(red.has_new_test_failures());
+
+        // Once the fix lands, the test passes again (green) — no new test failures.
+        let green = EvaluationResult::new(
+            1,
+            "org/repo".into(),
+            vec![EvalDelta::compute(
+                make_snapshot(EvalCategory::Test, "cargo test", 10, 0),
+                make_snapshot(EvalCategory::Test, "cargo test", 11, 0),
+            )],
+        );
+        assert!(!green.has_new_test_failures());
+
+        // A lint regression is not a test failure.
+        let lint = EvaluationResult::new(
+            1,
+            "org/repo".into(),
+            vec![EvalDelta::compute(
+                make_snapshot(EvalCategory::Lint, "clippy", 10, 0),
+                make_snapshot(EvalCategory::Lint, "clippy", 10, 1),
+            )],
+        );
+        assert!(!lint.has_new_test_failures());
+    }
+
     #[test]
     fn test_evaluation_result_pr_comment() {
         let before = make_snapshot(EvalCategory::Test, "cargo test", 10, 2);

crates/claudear-config/src/config.rs

@@ -805,6 +805,9 @@ pub struct ReplyConfig {
     pub templates: std::collections::HashMap<String, String>,
     /// Timeout for verifying (reproducing) a reported bug, in seconds (default: 1800).
     pub verify_timeout_secs: u64,
+    /// When verify can't run (timeout/error/unsupported), assume reproduced and fix
+    /// anyway (default: true). Set false to ask the reporter for repro steps instead.
+    pub verify_fail_open: bool,
 }
 
 impl Default for ReplyConfig {
@@ -815,6 +818,7 @@ impl Default for ReplyConfig {
             default_template: None,
             templates: std::collections::HashMap::new(),
             verify_timeout_secs: 1800,
+            verify_fail_open: true,
         }
     }
 }
@@ -1086,6 +1090,9 @@ pub struct EvaluationConfig {
     pub post_pr_comment: bool,
     /// Fail the fix attempt on regression.
     pub fail_on_regression: bool,
+    /// Enforce red->green: author a failing test first (must fail on the unfixed
+    /// code), then fix, then require it to pass. Needs test_delta enabled.
+    pub require_red_green: bool,
     /// Custom test command override.
     pub custom_test_cmd: Option<String>,
     /// Custom lint command override.
@@ -1108,6 +1115,7 @@ impl Default for EvaluationConfig {
             total_timeout_secs: 900,
             post_pr_comment: true,
             fail_on_regression: false,
+            require_red_green: false,
             custom_test_cmd: None,
             custom_lint_cmd: None,
             custom_analysis_cmd: None,
@@ -8120,6 +8128,7 @@ instructions_file = "my-instructions.md"
             total_timeout_secs: 1800,
             post_pr_comment: false,
             fail_on_regression: true,
+            require_red_green: false,
             custom_test_cmd: Some("npm test".to_string()),
             custom_lint_cmd: None,
             custom_analysis_cmd: Some("sonar".to_string()),

crates/claudear-core/src/types.rs

@@ -3232,6 +3232,18 @@ pub enum TimelineEventStatus {
     #[serde(rename = "verify_completed")]
     VerifyCompleted,
 
+    /// Red-green: the failing-test (red) phase began.
+    #[serde(rename = "red_green_started")]
+    RedGreenStarted,
+
+    /// Red-green: the authored test failed on the unfixed code (red confirmed).
+    #[serde(rename = "red_confirmed")]
+    RedConfirmed,
+
+    /// Red-green: the authored test passed after the fix (green confirmed).
+    #[serde(rename = "green_confirmed")]
+    GreenConfirmed,
+
     #[serde(rename = "reply_started")]
     ReplyStarted,
 
@@ -3292,6 +3304,9 @@ impl TimelineEventStatus {
             Self::FixStarted => "fix_started",
             Self::VerifyStarted => "verify_started",
             Self::VerifyCompleted => "verify_completed",
+            Self::RedGreenStarted => "red_green_started",
+            Self::RedConfirmed => "red_confirmed",
+            Self::GreenConfirmed => "green_confirmed",
             Self::ReplyStarted => "reply_started",
             Self::ReplySent => "reply_sent",
             Self::FixSucceeded => "fix_succeeded",

crates/claudear-e2e/src/config.rs

@@ -235,6 +235,7 @@ impl ConfigBuilder {
                 template.to_string(),
             )]),
             verify_timeout_secs,
+            verify_fail_open: true,
         };
         self
     }