feat: US-003 - Create sg-benchmarks mirrors for missing repos

Created 7 sg-benchmarks mirror repos on GitHub using orphan-commit approach:
- kubernetes-client-go (v12.0.0), kubernetes-api (v0.35.1)
- expressjs-express (v5.2.1), lodash (4.17.21)
- prisma-prisma (7.4.1), grafana-loki (v3.6.6), grafana-mimir (mimir-3.0.3)

All repos public, pinned to stable releases, SHAs recorded in
configs/sg_mirror_revisions.json. SG indexing pending (auto-crawl).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: sjarmak

ralph-mcp-unique/configs/sg_mirror_revisions.json

@@ -0,0 +1,70 @@
+{
+  "_description": "Tracks sg-benchmarks mirror repos: original source, pinned tag, original commit SHA, and mirror commit SHA. Used by repo-set fixtures to reference exact revisions.",
+  "_created": "2026-02-20",
+  "mirrors": [
+    {
+      "mirror_repo": "sg-benchmarks/kubernetes-client-go",
+      "source_repo": "kubernetes/client-go",
+      "tag": "v12.0.0",
+      "original_sha": "78d2af792babf2dd937ba2e2a8d99c753a5eda89",
+      "mirror_sha": "8020fc4fcf89965904a5f43689f169d6e01d1e80",
+      "indexed": false,
+      "created_date": "2026-02-20"
+    },
+    {
+      "mirror_repo": "sg-benchmarks/kubernetes-api",
+      "source_repo": "kubernetes/api",
+      "tag": "v0.35.1",
+      "original_sha": "4aa217d625944fe92a602d2ed6d3d28cc2748f0c",
+      "mirror_sha": "fa23dd302759dbb681c1a41f09d24190a38c1d58",
+      "indexed": false,
+      "created_date": "2026-02-20"
+    },
+    {
+      "mirror_repo": "sg-benchmarks/expressjs-express",
+      "source_repo": "expressjs/express",
+      "tag": "v5.2.1",
+      "original_sha": "dbac741a49a5a64336b70c06e85c2e2706e36336",
+      "mirror_sha": "9de5890d0dc6128b7f7eb15469d76aa60dacc48f",
+      "indexed": false,
+      "created_date": "2026-02-20",
+      "notes": "Clean-name mirror. Legacy mirror sg-benchmarks/expressjs--express--815f7993 also exists and is indexed."
+    },
+    {
+      "mirror_repo": "sg-benchmarks/lodash",
+      "source_repo": "lodash/lodash",
+      "tag": "4.17.21",
+      "original_sha": "f299b52f39486275a9e6483b60a410e06520c538",
+      "mirror_sha": "1dd1ecfd7875372efa8dde5dede50f6d2d323703",
+      "indexed": false,
+      "created_date": "2026-02-20"
+    },
+    {
+      "mirror_repo": "sg-benchmarks/prisma-prisma",
+      "source_repo": "prisma/prisma",
+      "tag": "7.4.1",
+      "original_sha": "533e22aa500fa87d6ac254b28f54a4446a1616c1",
+      "mirror_sha": "20117d718fb0db9c8a586276e2052f5b130f994b",
+      "indexed": false,
+      "created_date": "2026-02-20"
+    },
+    {
+      "mirror_repo": "sg-benchmarks/grafana-loki",
+      "source_repo": "grafana/loki",
+      "tag": "v3.6.6",
+      "original_sha": "fb202465c6e9fdda198e06d6588de8381ded79e7",
+      "mirror_sha": "a3af38d4da899032d3ee46a30932d072d37e1b9c",
+      "indexed": false,
+      "created_date": "2026-02-20"
+    },
+    {
+      "mirror_repo": "sg-benchmarks/grafana-mimir",
+      "source_repo": "grafana/mimir",
+      "tag": "mimir-3.0.3",
+      "original_sha": "0b00ebf5d3766c766d319a8c23cd165f5c60a5e5",
+      "mirror_sha": "cfaa8c9a705a2417822a5c6224a9fc3128c416c2",
+      "indexed": false,
+      "created_date": "2026-02-20"
+    }
+  ]
+}

ralph-mcp-unique/progress.txt

@@ -57,3 +57,29 @@
   - sourcegraph_mirror is the critical field linking to sg-benchmarks mirrors
   - access_mode on each repo entry is the authoritative source; local_checkout_repos/mcp_only_repos are convenience arrays
 ---
+[2026-02-20 19:50:26 UTC] Iteration 1 no story markers found
+[2026-02-20 19:50:26 UTC] Iteration 1 complete
+[2026-02-20 19:50:28 UTC] Iteration 2 started
+
+## 2026-02-20 - US-003: Create sg-benchmarks mirrors for missing repos
+- Created 7 sg-benchmarks mirror repos on GitHub using orphan-commit approach:
+  - sg-benchmarks/kubernetes-client-go (kubernetes/client-go @ v12.0.0)
+  - sg-benchmarks/kubernetes-api (kubernetes/api @ v0.35.1)
+  - sg-benchmarks/expressjs-express (expressjs/express @ v5.2.1)
+  - sg-benchmarks/lodash (lodash/lodash @ 4.17.21)
+  - sg-benchmarks/prisma-prisma (prisma/prisma @ 7.4.1)
+  - sg-benchmarks/grafana-loki (grafana/loki @ v3.6.6)
+  - sg-benchmarks/grafana-mimir (grafana/mimir @ mimir-3.0.3)
+- All repos public, main branch, orphan commit (no history, minimal push size)
+- Created `ralph-mcp-unique/configs/sg_mirror_revisions.json` tracking all SHAs
+- Created `ralph-mcp-unique/scripts/verify_sg_indexing.py` to check SG indexing status
+- Legacy express mirror `expressjs--express--815f7993` also exists (indexed)
+- **PENDING**: Sourcegraph indexing — repos created but not yet indexed. SG auto-crawls sg-benchmarks org; indexing expected within hours. Run verify_sg_indexing.py to check.
+- Files changed: `ralph-mcp-unique/configs/sg_mirror_revisions.json` (new), `ralph-mcp-unique/scripts/verify_sg_indexing.py` (new)
+- **Learnings for future iterations:**
+  - `gh api repos/sg-benchmarks/<name>` returns 404 for newly created repos; cannot use as existence check before the API propagates
+  - Orphan-commit approach works for all repo sizes (even 500MB+ repos like loki/mimir push fine)
+  - Clean naming convention (org-repo) preferred over SWE-bench hash-suffix convention (org--repo--hash)
+  - SG indexing delay is the bottleneck — create mirrors early, verify later
+  - `.gitignore` has `ralph-*/` pattern; must use `git add -f` for new files under ralph-mcp-unique/
+---

ralph-mcp-unique/scripts/verify_sg_indexing.py

@@ -0,0 +1,87 @@
+#!/usr/bin/env python3
+"""Verify that sg-benchmarks mirrors are indexed in Sourcegraph.
+
+Run periodically after creating mirrors until all are indexed.
+Uses the Sourcegraph GraphQL API via the `src` CLI or direct HTTP.
+
+Usage:
+    python3 scripts/verify_sg_indexing.py [--update]
+
+    --update: Update configs/sg_mirror_revisions.json indexed field
+"""
+import json
+import subprocess
+import sys
+from pathlib import Path
+
+REVISIONS_FILE = Path(__file__).parent.parent / "configs" / "sg_mirror_revisions.json"
+
+
+def check_repo_indexed(mirror_repo: str) -> bool:
+    """Check if a repo is indexed in Sourcegraph using keyword_search."""
+    # Use src CLI if available, otherwise use gh + SG API
+    # For now, check if repo appears in SG search results
+    try:
+        # Try a simple search for any file in the repo
+        result = subprocess.run(
+            ["src", "search", "-json", f"repo:^github.com/{mirror_repo}$ count:1 file:."],
+            capture_output=True, text=True, timeout=30
+        )
+        if result.returncode == 0:
+            data = json.loads(result.stdout)
+            return len(data.get("Results", [])) > 0
+    except (FileNotFoundError, subprocess.TimeoutExpired, json.JSONDecodeError):
+        pass
+
+    # Fallback: check via GitHub API if SG has cloned the repo
+    # (This is a heuristic - repo existence on GH doesn't mean SG indexed it)
+    print(f"  Note: `src` CLI not available. Cannot verify SG indexing for {mirror_repo}.")
+    print(f"  Manually verify: search 'repo:^github.com/{mirror_repo}$' on Sourcegraph.")
+    return False
+
+
+def main():
+    update = "--update" in sys.argv
+
+    if not REVISIONS_FILE.exists():
+        print(f"ERROR: {REVISIONS_FILE} not found")
+        sys.exit(1)
+
+    data = json.loads(REVISIONS_FILE.read_text())
+    mirrors = data["mirrors"]
+
+    all_indexed = True
+    results = []
+
+    for mirror in mirrors:
+        repo = mirror["mirror_repo"]
+        was_indexed = mirror.get("indexed", False)
+
+        if was_indexed:
+            status = "INDEXED (previously verified)"
+            indexed = True
+        else:
+            indexed = check_repo_indexed(repo)
+            status = "INDEXED" if indexed else "NOT YET INDEXED"
+
+        if not indexed:
+            all_indexed = False
+
+        results.append((repo, status, indexed))
+        print(f"  {repo}: {status}")
+
+        if update and indexed and not was_indexed:
+            mirror["indexed"] = True
+
+    if update:
+        REVISIONS_FILE.write_text(json.dumps(data, indent=2) + "\n")
+        print(f"\nUpdated {REVISIONS_FILE}")
+
+    print(f"\n{'ALL INDEXED' if all_indexed else 'SOME NOT YET INDEXED'}")
+    print(f"Indexed: {sum(1 for _, _, i in results if i)}/{len(results)}")
+
+    sys.exit(0 if all_indexed else 1)
+
+
+if __name__ == "__main__":
+    main()