chore: cleanup dead code, add 9 EP benchmark config

- Remove unused _build_improvement_prompt, improve_tool, run_feedback_loop
- Fix mislabeled "missed files" line in improvement prompt
- Remove unused imports
- Add config.yaml with 9 EPs across 4 operators
- EP #1964 excluded (openshift/must-gather is Shell, not Go)

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: rausingh-rh

benchmark/benchmark.py

@@ -161,9 +161,14 @@ async def cmd_measure(args: argparse.Namespace) -> None:
     results: list[BenchmarkResult] = []
     for i, case in enumerate(config.cases, 1):
         logger.info("\n>>> EP %d of %d <<<", i, len(config.cases))
-        result = await run_single_ep(case, config, output_dir, "measure", force=args.force)
-        if result:
-            results.append(result)
+        try:
+            result = await run_single_ep(case, config, output_dir, "measure", force=args.force)
+            if result:
+                results.append(result)
+        except Exception:
+            ep_num = _ep_num(case.ep_url)
+            logger.error("EP #%s failed, skipping: %s", ep_num, __import__("traceback").format_exc())
+            continue
 
     if results:
         logger.info("\n" + "=" * 60)
@@ -250,9 +255,14 @@ async def cmd_verify(args: argparse.Namespace) -> None:
     results: list[BenchmarkResult] = []
     for i, case in enumerate(config.cases, 1):
         logger.info("\n>>> EP %d of %d <<<", i, len(config.cases))
-        result = await run_single_ep(case, config, output_dir, "verify", force=args.force)
-        if result:
-            results.append(result)
+        try:
+            result = await run_single_ep(case, config, output_dir, "verify", force=args.force)
+            if result:
+                results.append(result)
+        except Exception:
+            ep_num = _ep_num(case.ep_url)
+            logger.error("EP #%s failed, skipping: %s", ep_num, __import__("traceback").format_exc())
+            continue
 
     if results:
         logger.info("\n" + "=" * 60)

benchmark/config.yaml

@@ -1,15 +1,52 @@
 benchmark_cases:
+  - ep_url: "https://github.com/openshift/enhancements/pull/1863"
+    repo_url: "https://github.com/openshift/zero-trust-workload-identity-manager"
+    description: "SPIRE federation support"
+    implementation_prs: [68, 82]
+
   - ep_url: "https://github.com/openshift/enhancements/pull/1834"
     repo_url: "https://github.com/openshift/external-secrets-operator"
-    description: "External Secrets Operator enhancement"
+    description: "Network Policy for ESO"
     implementation_prs: [67, 74]
 
+  - ep_url: "https://github.com/openshift/enhancements/pull/1898"
+    repo_url: "https://github.com/openshift/external-secrets-operator"
+    description: "ESO install-time customizations"
+    implementation_prs: [91, 94, 97, 106, 111]
+
+  - ep_url: "https://github.com/openshift/enhancements/pull/1914"
+    repo_url: "https://github.com/openshift/cert-manager-operator"
+    description: "TrustManager support in cert-manager"
+    implementation_prs: [362, 371, 379]
+
+  - ep_url: "https://github.com/openshift/enhancements/pull/1923"
+    repo_url: "https://github.com/openshift/must-gather-operator"
+    description: "since/sinceTime fields in MustGather CR"
+    implementation_prs: [323]
+
+  - ep_url: "https://github.com/openshift/enhancements/pull/1903"
+    repo_url: "https://github.com/openshift/must-gather-operator"
+    description: "Remove proxy config from MustGather CR"
+    implementation_prs: [313]
+
+  - ep_url: "https://github.com/openshift/enhancements/pull/1824"
+    repo_url: "https://github.com/openshift/zero-trust-workload-identity-manager"
+    description: "OIDC discovery provider route"
+    implementation_prs: [33]
+
+  - ep_url: "https://github.com/openshift/enhancements/pull/1906"
+    repo_url: "https://github.com/openshift/must-gather-operator"
+    description: "Custom image option in MustGather spec"
+    implementation_prs: [322]
+
+  - ep_url: "https://github.com/openshift/enhancements/pull/1839"
+    repo_url: "https://github.com/openshift/must-gather-operator"
+    description: "PV to persist must-gather data"
+    implementation_prs: [294]
+
+# EP #1964 excluded: openshift/must-gather is Shell-based, not a Go operator
+
 settings:
-  tools_to_benchmark:
-    - api-generate
-    - api-implement
-  iterations: 3
-  output_dir: "benchmark/results"
-  parallel: false
   model: "claude-opus-4-6"
   effort: "max"
+  output_dir: "benchmark/results"

benchmark/isolate.py

@@ -15,9 +15,20 @@
 logger = logging.getLogger(__name__)
 
 
-def _run(cmd: list[str], **kwargs) -> subprocess.CompletedProcess:
+def _run(cmd: list[str], retries: int = 3, **kwargs) -> subprocess.CompletedProcess:
     logger.debug("Running: %s", " ".join(cmd))
-    return subprocess.run(cmd, capture_output=True, text=True, check=True, **kwargs)
+    for attempt in range(retries):
+        result = subprocess.run(cmd, capture_output=True, text=True, **kwargs)
+        if result.returncode == 0:
+            return result
+        if attempt < retries - 1:
+            import time
+            wait = 5 * (attempt + 1)
+            logger.warning("Command failed (attempt %d/%d), retrying in %ds: %s",
+                           attempt + 1, retries, wait, result.stderr[:200])
+            time.sleep(wait)
+    result.check_returncode()
+    return result
 
 
 def _gh_json(args: list[str]) -> dict | list: