Add cross-task content methods, visuals, and scoring

Why: Layout rules alone do not produce conclusion-first documents, extra views, or infographics, and nothing scored a deliverable before it was returned.
Evidence: go test ./internal/prompts/ ./internal/quality/ ./internal/tools/ ./internal/squadadapter/ ./eval/ -count=1 and the house-style unittest passed in the previous session.
Impact: Every task inherits content rules. Meetings, reports, and comparisons can name a view family. Long documents without a visual summary warn. quality_review and the squad checker share one scorecard.
Refs: deliverable-quality-system A1-A5
Co-authored-by: Cursor <cursoragent@cursor.com>

eval/README.md

@@ -0,0 +1,9 @@
+# 交付物评测
+
+仓内 `testdata/` 是脱敏短样例，用来确认评分卡能区分“结论先行”和“栏目式笔记”。
+
+真实会议材料不要放进仓库。放到 `~/.agent-system/eval/`，或设置 `EVAL_DIR`。其中的 `.md` 会在 `go test ./eval/` 里被打分并打印，缺失该目录时测试跳过。
+
+```
+cd modules/backend && go test ./eval/ ./internal/quality/ -count=1
+```

eval/compare.go

@@ -0,0 +1,14 @@
+package eval
+
+import "strings"
+
+// MissingThemes 报告候选文本没有覆盖的参考主题。主题是短词，不是整句照抄。
+func MissingThemes(candidate string, themes []string) []string {
+	var missing []string
+	for _, theme := range themes {
+		if !strings.Contains(candidate, theme) {
+			missing = append(missing, theme)
+		}
+	}
+	return missing
+}

eval/deliverable_test.go

@@ -0,0 +1,62 @@
+// Package eval 用固定样例比较交付质量。真实会议材料放在仓外
+// ~/.agent-system/eval/，或用 EVAL_DIR 指向该目录；仓内只放脱敏短样例。
+package eval
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"agent-system-backend/internal/quality"
+)
+
+func TestDesensitizedSampleBeatsColumnNotes(t *testing.T) {
+	good := scoreFixture(t, "good.md")
+	bad := scoreFixture(t, "bad.md")
+	if !good.Passed {
+		t.Fatalf("good fixture should pass: %+v", good)
+	}
+	if bad.Score >= good.Score {
+		t.Fatalf("bad %d should be below good %d", bad.Score, good.Score)
+	}
+	missing := MissingThemes(readFixture(t, "good.md"), []string{"相关性", "标签", "未指明"})
+	if len(missing) > 0 {
+		t.Fatalf("good fixture missed themes %v", missing)
+	}
+}
+
+func TestExternalEvalDirOptional(t *testing.T) {
+	dir := os.Getenv("EVAL_DIR")
+	if dir == "" {
+		home, err := os.UserHomeDir()
+		if err != nil {
+			t.Skip(err)
+		}
+		dir = filepath.Join(home, ".agent-system", "eval")
+	}
+	entries, err := os.ReadDir(dir)
+	if err != nil {
+		t.Skip("no external eval dir")
+	}
+	for _, entry := range entries {
+		if entry.IsDir() || filepath.Ext(entry.Name()) != ".md" {
+			continue
+		}
+		report := quality.ScoreFile(filepath.Join(dir, entry.Name()))
+		t.Logf("%s score=%d passed=%v", entry.Name(), report.Score, report.Passed)
+	}
+}
+
+func scoreFixture(t *testing.T, name string) quality.Report {
+	t.Helper()
+	return quality.ScoreText(readFixture(t, name))
+}
+
+func readFixture(t *testing.T, name string) string {
+	t.Helper()
+	data, err := os.ReadFile(filepath.Join("testdata", name))
+	if err != nil {
+		t.Fatal(err)
+	}
+	return string(data)
+}

eval/testdata/bad.md

@@ -0,0 +1,3 @@
+# 讨论过程
+
+今天围绕文档上传聊了一会儿。大家说了很多，也有一些不同看法。后续再看吧。

eval/testdata/good.md

@@ -0,0 +1,16 @@
+# 上传文档应先判断相关性，再打标签入库
+
+## 先校验，再生成
+说话人1指出，上传后的第一步是判断文档与项目是否相关。确认后再拆开打标签。
+一句话：没有这层记忆，后面的摘要和追溯都没有依据。
+
+## 不要在上传时提示缺什么
+依据会议原话，缺项提示放到体系成型之后。
+一句话：上传节点是收集，不是验收。
+
+## 基线要能迭代
+新内容和已有基线比对，冲突时先问人。
+一句话：一份基线反复补，而不是每次上传都新开一篇。
+
+## 下一步
+事项是写出第一版需求。负责人是陈威。期限是周五前。说不清的一律写未指明。