perf(edit_apply): drop redundant Diff: label; docs(file_read): note strip renumbering

edit_apply response=diff prefixed the fenced ```diff block with a 'Diff:'
label that the fence and the diff's own ---/+++ headers already convey — pure
token cost per diff response. Extract fencedDiff() (also dedups the diff and
default branches) and drop the label. Adds a test guard.

Also document that strip renumbers lines from 1 over the stripped content, so
the model doesn't cite original-file line numbers after a stripped read.

internal/server/features_test.go

@@ -215,6 +215,36 @@ func TestASTQuerySkipsDependencyDirs(t *testing.T) {
 
 // ---- Feature 3 (ast_query): pagination cursor ----
 
+// TestASTQueryFirstPageCursor is a strict regression guard: page 1 must emit a
+// cursor with an accurate remaining count when more matches exist. Previously
+// the page-1 walk was capped at max_results, so the total was truncated and the
+// cursor fired nondeterministically (and with a wrong remaining count).
+func TestASTQueryFirstPageCursor(t *testing.T) {
+	srv, tmpDir := newFeaturesServer(t)
+	ctx := context.Background()
+
+	// Fixture has 5 functions (Alpha/Beta/Gamma/Delta/Epsilon); cap at 2.
+	req := mcp.CallToolRequest{}
+	req.Params.Arguments = map[string]interface{}{
+		"pattern":     "func $NAME() $RET",
+		"language":    "go",
+		"paths":       []interface{}{tmpDir},
+		"max_results": float64(2),
+	}
+	res, err := srv.handleASTQuery(ctx, req)
+	if err != nil {
+		t.Fatalf("page1 error: %v", err)
+	}
+	text := res.Content[0].(mcp.TextContent).Text
+	if !strings.Contains(text, "[cursor:") {
+		t.Fatalf("page1 with 5 matches capped at 2 must emit a cursor, got:\n%s", text)
+	}
+	// 5 total - 2 shown = 3 remaining.
+	if !strings.Contains(text, "remaining: 3") {
+		t.Errorf("expected accurate 'remaining: 3', got:\n%s", text)
+	}
+}
+
 func TestASTQueryPaginationCursor(t *testing.T) {
 	srv, tmpDir := newFeaturesServer(t)
 	ctx := context.Background()

internal/server/handlers_ast.go

@@ -109,12 +109,13 @@ func (s *Server) runASTQueryWalk(ctx context.Context, p *astQueryParams, exts []
 		},
 	}
 
-	// Collect limit for early-exit: when paginating we need all results first.
-	collectLimit := p.maxResults
-	if p.offset > 0 {
-		collectLimit = 0
-	}
-
+	// A full walk is required on every page: the cursor's "remaining" count is
+	// derived from the total match count (see buildASTCursorFooter), so the page-1
+	// response can only emit a correct cursor if it has seen every match. Capping
+	// page 1 at max_results made the cursor fire nondeterministically (only when
+	// the last parsed file overshot the cap) with a wrong remaining count. Parse
+	// results are LRU-cached, so re-walking on later pages is mostly file I/O, and
+	// matches below max_results already visit the whole tree regardless.
 	var allResults []query.MatchResult
 	for _, searchPath := range p.paths {
 		if !s.cfg.IsPathAllowed(searchPath) {
@@ -162,10 +163,6 @@ func (s *Server) runASTQueryWalk(ctx context.Context, p *astQueryParams, exts []
 				return nil
 			}
 			allResults = append(allResults, matches...)
-
-			if collectLimit > 0 && len(allResults) >= collectLimit {
-				return filepath.SkipAll
-			}
 			return nil
 		})
 		if walkErr != nil {

internal/server/handlers_edit.go

@@ -92,22 +92,21 @@ func (s *Server) handleEdit(ctx context.Context, request mcp.CallToolRequest) (*
 		return mcp.NewToolResultText(fmt.Sprintf("+%d -%d", added, removed)), nil
 
 	case "diff":
-		var output strings.Builder
-		output.WriteString("Diff:\n```diff\n")
-		output.WriteString(result.Diff)
-		output.WriteString("```\n")
-		return mcp.NewToolResultText(output.String()), nil
+		return mcp.NewToolResultText(fencedDiff(result.Diff)), nil
 
 	default:
 		// Fallback: treat unknown values as "diff" for safety.
-		var output strings.Builder
-		output.WriteString("Diff:\n```diff\n")
-		output.WriteString(result.Diff)
-		output.WriteString("```\n")
-		return mcp.NewToolResultText(output.String()), nil
+		return mcp.NewToolResultText(fencedDiff(result.Diff)), nil
 	}
 }
 
+// fencedDiff wraps a unified diff in a ```diff code fence. No "Diff:" label is
+// added — the fence and the diff's own ---/+++ headers already identify it, so
+// the label would only cost tokens.
+func fencedDiff(diff string) string {
+	return "```diff\n" + diff + "```\n"
+}
+
 // countDiffLines counts added (+) and removed (-) content lines in a unified diff.
 // Only lines inside hunks are counted (everything after the first "@@" header), so the
 // "---"/"+++" file headers are skipped structurally — and content whose own text starts

internal/server/handlers_edit_test.go

@@ -59,6 +59,33 @@ func TestEditApplyPreservesBackslashSequencesText(t *testing.T) {
 	}
 }
 
+// TestEditApplyDiffResponseFencedNoLabel verifies response=diff emits the unified
+// diff in a ```diff fence with no redundant "Diff:" label (the fence and the diff's
+// own headers already identify it; the label would only cost tokens).
+func TestEditApplyDiffResponseFencedNoLabel(t *testing.T) {
+	tmpDir := t.TempDir()
+	srv := newTestServer(t, tmpDir)
+	f := writeFile(t, tmpDir, "note.txt", "before\n")
+
+	res := callEdit(t, srv, map[string]any{
+		"file":          f,
+		"operation":     "replace",
+		"selector_text": "before",
+		"new_content":   "after",
+		"response":      "diff",
+	})
+	if res.IsError {
+		t.Fatalf("edit returned error: %+v", res.Content)
+	}
+	out := res.Content[0].(mcp.TextContent).Text
+	if !strings.HasPrefix(out, "```diff\n") {
+		t.Errorf("diff response should start with the ```diff fence, got:\n%s", out)
+	}
+	if strings.Contains(out, "Diff:") {
+		t.Errorf("diff response should not carry a redundant 'Diff:' label, got:\n%s", out)
+	}
+}
+
 // TestCountDiffLinesHunkAware verifies the line counter skips file headers structurally and
 // counts content lines whose own text begins with + or - (previously dropped as "+++"/"---").
 func TestCountDiffLinesHunkAware(t *testing.T) {

internal/server/server.go

@@ -193,7 +193,7 @@ func (s *Server) registerTools() {
 				mcp.Description("Output mode: 'full' (default, full file), 'skeleton' (signatures + { ... } stubs, bodies elided), 'symbols_only' (symbol list only, alias for symbols_only=true)."),
 			),
 			mcp.WithArray("strip",
-				mcp.Description("Strip content classes before line-numbering. Values: 'imports' (remove import blocks), 'license' (remove leading license comment), 'block_comments' (remove /* */ and Python triple-quoted strings). Emits [stripped: ...] footer."),
+				mcp.Description("Strip content classes before line-numbering. Values: 'imports' (remove import blocks), 'license' (remove leading license comment), 'block_comments' (remove /* */ and Python triple-quoted strings). Emits [stripped: ...] footer. Note: line numbers are renumbered from 1 over the stripped content, so they no longer match the file's original line numbers."),
 				mcp.WithStringItems(),
 			),
 			mcp.WithNumber("max_lines",