fix(prompt): separate review findings

internal/daemon/hooks_test.go

@@ -1039,9 +1039,20 @@ func TestHandleEventNoLogWhenNoHooksMatch(t *testing.T) {
 }
 
 func TestWaitUntilIdle_ConcurrentEvents(t *testing.T) {
-	t.Parallel()
 	// A dedicated stress test proving WaitUntilIdle waits past the event-processing boundary
 	// under timing races and concurrent broadcasts.
+	if runtime.GOOS != "windows" {
+		t.Parallel()
+	}
+
+	iterations := 50
+	numEvents := 10
+	if runtime.GOOS == "windows" {
+		// Each hook starts a PowerShell process on Windows. Keep the test broad
+		// enough to exercise concurrent broadcasts without exhausting CI time.
+		iterations = 5
+		numEvents = 4
+	}
 
 	tmpDir := t.TempDir()
 
@@ -1053,11 +1064,10 @@ func TestWaitUntilIdle_ConcurrentEvents(t *testing.T) {
 		},
 	}
 
-	for i := range 50 {
+	for i := range iterations {
 		hr, broadcaster := setupRunner(t, cfg)
 
 		var wg sync.WaitGroup
-		numEvents := 10
 
 		for j := range numEvents {
 			wg.Add(1)

internal/prompt/templates/claude-code_review.md.gotmpl

@@ -11,6 +11,7 @@ Do NOT build the project, run the test suite, or execute the code while reviewin
   - **Location**: File and line number where possible
   - **Problem**: Concise description of the issue
   - **Fix**: Brief suggested fix
+- Separate multiple findings with `---` on its own line.
 - After the findings, add `## Summary` with a single sentence summarizing the change.
 - If you find no issues, start immediately with `No issues found.` and then add `Summary: <one sentence>`.
 

internal/prompt/templates/codex_review.md.gotmpl

@@ -12,6 +12,7 @@ Do NOT search or read files outside the repository checkout, except for an expli
   - **Location**: File and line number where possible
   - **Problem**: Concise description of the issue
   - **Fix**: Brief suggested fix
+- Separate multiple findings with `---` on its own line.
 - After the findings, add `## Summary` with a single sentence summarizing the change.
 - If you find no issues, start immediately with `No issues found.` and then add `Summary: <one sentence>`.
 

internal/prompt/templates/default_dirty.md.gotmpl

@@ -23,6 +23,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.
 

internal/prompt/templates/default_range.md.gotmpl

@@ -28,6 +28,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.
 

internal/prompt/templates/default_review.md.gotmpl

@@ -28,6 +28,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.
 

internal/prompt/templates/default_security.md.gotmpl

@@ -24,6 +24,7 @@ For each finding, provide:
 - File and line reference
 - The specific code path an attacker would exploit and what they gain
 - Suggested remediation
+- Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file-level when the issue spans a range) and describe a plausible exploit path. The severity must match the exploitability you described. Drop any finding that fails these checks.
 

internal/prompt/templates/gemini_review.md.gotmpl

@@ -13,6 +13,7 @@ Do NOT build the project, run the test suite, or execute the code while reviewin
      - **Location**: File and line number
      - **Problem**: Concise description
      - **Fix**: Brief suggested fix
+   - Separate multiple findings with `---` on its own line.
    - If no issues are found, state "No issues found."
 
 ## Review Criteria

internal/prompt/templates_test.go

@@ -109,6 +109,33 @@ func TestGetSystemPrompt_Fallbacks(t *testing.T) {
 	}
 }
 
+func TestGetSystemPrompt_ReviewPromptsRequireFindingSeparators(t *testing.T) {
+	fixedTime := time.Date(2030, 6, 15, 0, 0, 0, 0, time.UTC)
+	mockNow := func() time.Time { return fixedTime }
+
+	tests := []struct {
+		name    string
+		agent   string
+		command string
+	}{
+		{name: "default review", agent: "test", command: "review"},
+		{name: "default dirty", agent: "test", command: "dirty"},
+		{name: "default range", agent: "test", command: "range"},
+		{name: "default security", agent: "test", command: "security"},
+		{name: "claude review", agent: "claude-code", command: "review"},
+		{name: "codex review", agent: "codex", command: "review"},
+		{name: "gemini review", agent: "gemini", command: "review"},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got := getSystemPrompt(tt.agent, tt.command, mockNow)
+
+			assert.Contains(t, got, "Separate multiple findings with `---` on its own line.")
+		})
+	}
+}
+
 func TestGetSystemPrompt_DateInjection(t *testing.T) {
 	fixedTime := time.Date(2030, 6, 15, 0, 0, 0, 0, time.UTC)
 	fixedDateStr := "Current date: 2030-06-15 (UTC)"

internal/prompt/testdata/golden/dirty_review.golden

@@ -23,6 +23,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.
 

internal/prompt/testdata/golden/dirty_truncated.golden

@@ -23,6 +23,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.
 
@@ -112,9 +113,6 @@ index 0000000..1111111
 +a line of content
 +a line of content
 +a line of content
-+a line of content
-+a line of content
-+a line of content
-+a line of conte
++a line of co
 ... (truncated)
 ```

internal/prompt/testdata/golden/previous_reviews_with_comments.golden

@@ -28,6 +28,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.
 

internal/prompt/testdata/golden/range_truncated.golden

@@ -28,6 +28,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.
 

internal/prompt/testdata/golden/range_truncated_codex_in_range.golden

@@ -12,6 +12,7 @@ Do NOT search or read files outside the repository checkout, except for an expli
   - **Location**: File and line number where possible
   - **Problem**: Concise description of the issue
   - **Fix**: Brief suggested fix
+- Separate multiple findings with `---` on its own line.
 - After the findings, add `## Summary` with a single sentence summarizing the change.
 - If you find no issues, start immediately with `No issues found.` and then add `Summary: <one sentence>`.
 

internal/prompt/testdata/golden/range_with_in_range_reviews.golden

@@ -28,6 +28,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.
 

internal/prompt/testdata/golden/security_review.golden

@@ -24,6 +24,7 @@ For each finding, provide:
 - File and line reference
 - The specific code path an attacker would exploit and what they gain
 - Suggested remediation
+- Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file-level when the issue spans a range) and describe a plausible exploit path. The severity must match the exploitability you described. Drop any finding that fails these checks.
 

internal/prompt/testdata/golden/single_review_claude_code.golden

@@ -11,6 +11,7 @@ Do NOT build the project, run the test suite, or execute the code while reviewin
   - **Location**: File and line number where possible
   - **Problem**: Concise description of the issue
   - **Fix**: Brief suggested fix
+- Separate multiple findings with `---` on its own line.
 - After the findings, add `## Summary` with a single sentence summarizing the change.
 - If you find no issues, start immediately with `No issues found.` and then add `Summary: <one sentence>`.
 

internal/prompt/testdata/golden/single_review_codex.golden

@@ -12,6 +12,7 @@ Do NOT search or read files outside the repository checkout, except for an expli
   - **Location**: File and line number where possible
   - **Problem**: Concise description of the issue
   - **Fix**: Brief suggested fix
+- Separate multiple findings with `---` on its own line.
 - After the findings, add `## Summary` with a single sentence summarizing the change.
 - If you find no issues, start immediately with `No issues found.` and then add `Summary: <one sentence>`.
 

internal/prompt/testdata/golden/single_review_default.golden

@@ -28,6 +28,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.
 

internal/prompt/testdata/golden/single_review_gemini.golden

@@ -13,6 +13,7 @@ Do NOT build the project, run the test suite, or execute the code while reviewin
      - **Location**: File and line number
      - **Problem**: Concise description
      - **Fix**: Brief suggested fix
+   - Separate multiple findings with `---` on its own line.
    - If no issues are found, state "No issues found."
 
 ## Review Criteria

internal/prompt/testdata/golden/single_truncated_diff.golden

@@ -28,6 +28,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.
 

internal/prompt/testdata/golden/single_truncated_diff_codex.golden

@@ -12,6 +12,7 @@ Do NOT search or read files outside the repository checkout, except for an expli
   - **Location**: File and line number where possible
   - **Problem**: Concise description of the issue
   - **Fix**: Brief suggested fix
+- Separate multiple findings with `---` on its own line.
 - After the findings, add `## Summary` with a single sentence summarizing the change.
 - If you find no issues, start immediately with `No issues found.` and then add `Summary: <one sentence>`.
 

internal/prompt/testdata/golden/single_with_additional_context.golden

@@ -28,6 +28,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.
 

internal/prompt/testdata/golden/single_with_guidelines.golden

@@ -28,6 +28,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.
 

internal/prompt/testdata/golden/single_with_previous_reviews.golden

@@ -28,6 +28,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.
 

internal/prompt/testdata/golden/single_with_severity_filter.golden

@@ -28,6 +28,7 @@ After reviewing, provide:
    - File and line reference where possible
    - What specifically goes wrong if this is not fixed (concrete harm, not "violates best practices")
    - Suggested fix
+   - Separate multiple findings with `---` on its own line.
 
 Before finalizing, verify your review: every finding must reference the narrowest applicable location (line number when possible, file or diff-level when the issue is an omission or spans a range), the severity must match the impact you described, and no two findings should contradict each other. Drop any finding that fails these checks.