Add eval fixture pack, rule-level metrics, and PR rule-hit analytics

Author: haasonsaas

README.md

@@ -133,13 +133,17 @@ expect:
       contains: missing auth check
       severity: error
       category: security
-      rule_id: sec.auth.guard
+      rule_id: sec.auth.guard      # label for per-rule precision/recall
+      require_rule_id: false       # set true to require model-emitted RULE id
   must_not_find:
     - contains: style
   min_total: 1
   max_total: 8
 ```
 
+`diffscope eval` now reports per-rule precision/recall/F1 (micro and macro), and includes top rule-level TP/FP/FN counts in CLI and JSON output.
+Starter fixtures live in `eval/fixtures/repo_regressions`.
+
 ### Smart Review (Enhanced Analysis)
 ```bash
 # Get professional-grade analysis with confidence scoring

eval/fixtures/README.md

@@ -0,0 +1,17 @@
+# Eval Fixtures
+
+Starter fixture set for `diffscope eval`.
+
+- `repo_regressions/` contains regression-style diffs based on realistic mistakes in this codebase.
+- Each fixture can include `rule_id` as a label for rule-level precision/recall metrics.
+- Set `require_rule_id: true` on a pattern if the rule id must be emitted by the model for a match.
+
+Run:
+
+```bash
+diffscope eval --fixtures eval/fixtures --output eval-report.json
+```
+
+Notes:
+- Fixtures call the configured model and API provider; they are not deterministic unit tests.
+- Treat this set as a baseline and tighten `must_find`/`must_not_find` thresholds over time.

eval/fixtures/repo_regressions/raw_comment_missing_field.yml

@@ -0,0 +1,10 @@
+name: repo regression - missing RawComment rule_id initializer
+repo_path: ../../..
+diff_file: ./raw_comment_missing_field.patch
+expect:
+  must_find:
+    - file: src/main.rs
+      contains: rule_id
+      rule_id: compile.rawcomment.rule_id
+  min_total: 1
+  max_total: 8

eval/fixtures/repo_regressions/shell_injection.yml

@@ -0,0 +1,12 @@
+name: repo regression - shell injection in debug helper
+repo_path: ../../..
+diff_file: ./shell_injection.patch
+expect:
+  must_find:
+    - file: src/main.rs
+      contains: injection
+      rule_id: sec.shell.injection
+  must_not_find:
+    - contains: style
+  min_total: 1
+  max_total: 12

eval/fixtures/repo_regressions/smart_confidence_unwrap.yml

@@ -0,0 +1,12 @@
+name: repo regression - unwrap panic in confidence parser
+repo_path: ../../..
+diff_file: ./smart_confidence_unwrap.patch
+expect:
+  must_find:
+    - file: src/main.rs
+      contains: unwrap
+      rule_id: reliability.unwrap_panic
+  must_not_find:
+    - contains: style
+  min_total: 1
+  max_total: 10