fix: paired baseline+MCP execution and correct repo names in instructions

Runner (run_selected_tasks.sh):
- Replace mode-sequential passes with per-task paired execution (_launch_task_pair)
  matching _common.sh's run_paired_configs pattern: baseline + MCP start simultaneously
  per task so timing is comparable and resource utilization is maximized
- Dockerfile swap/restore moved to task-pair level: swap once before launching both
  configs, background watcher restores after BOTH complete (no race condition)

Instructions (10 tasks):
- Replace generic "org/repo-name" placeholder with exact repo identifiers the oracle
  uses (e.g. "sg-benchmarks/kubernetes-client-go", "sg-benchmarks/grafana-loki")
- Add "Important: Use exact repo identifiers" note to all affected tasks
- Fix stale commit SHAs: 8020fc4 → v0.32.0 (client-go), a3af38d → v3.3.4 (loki)
- Root cause of 0.0 scores: baseline agents wrote GitHub repo names instead of
  sg-benchmarks mirror names that the oracle file_set_match check expects

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

Author: sjarmak

benchmarks/ccb_mcp_crossorg/ccx-crossorg-061/instruction.md

@@ -38,7 +38,7 @@ Create a file at `/workspace/answer.json` with your findings:
 {
   "symbols": [
     {
-      "repo": "org/repo-name",
+      "repo": "kubernetes/kubernetes",
       "path": "relative/path/to/file.go",
       "symbol": "StructName"
     }
@@ -47,6 +47,8 @@ Create a file at `/workspace/answer.json` with your findings:
 }
 ```
 
+**Important**: Use exact repo identifiers as they appear in Sourcegraph. The oracle expects entries for `kubernetes/kubernetes` and `grafana/grafana`. The `repo` field must match these exactly.
+
 ## Evaluation
 
 Your answer is evaluated on:

benchmarks/ccb_mcp_crossrepo_tracing/ccx-config-trace-010/instruction.md

@@ -28,7 +28,7 @@ directories, but the authoritative source lives in separate repositories accessi
 
 The local `/workspace/` directory contains all repositories:
 - `kubernetes/kubernetes` at v1.32.0 → `/workspace/kubernetes`
-- `kubernetes/client-go` at 8020fc4 → `/workspace/client-go`
+- `kubernetes/client-go` at v0.32.0 → `/workspace/client-go`
 - `kubernetes/api` at fa23dd3 → `/workspace/api`
 - `etcd-io/etcd` at v3.5.17 → `/workspace/etcd`
 
@@ -39,12 +39,14 @@ Create a file at `/workspace/answer.json` with your findings in the following st
 ```json
 {
   "symbols": [
-    {"repo": "org/repo-name", "path": "relative/path/to/file.go", "symbol": "SymbolName"}
+    {"repo": "sg-benchmarks/kubernetes-client-go", "path": "relative/path/to/file.go", "symbol": "SymbolName"}
   ],
   "text": "Explanation of where Config is defined, the package import path, and why this is the authoritative source."
 }
 ```
 
+**Important**: The local `/workspace/client-go` directory contains the `kubernetes/client-go` source, but in Sourcegraph it is indexed as `sg-benchmarks/kubernetes-client-go`. Use `sg-benchmarks/kubernetes-client-go` as the `repo` value in your answer — the oracle checks for this exact identifier.
+
 Your answer is evaluated against a closed-world oracle — the exact repo, path, and symbol name matter.
 
 ## Evaluation

benchmarks/ccb_mcp_crossrepo_tracing/ccx-dep-trace-001/instruction.md

@@ -22,7 +22,7 @@ that directly imports it (not just subpackages) will be affected by a breaking A
 
 The local `/workspace/` directory contains all repositories:
 - `kubernetes/kubernetes` at v1.32.0 → `/workspace/kubernetes`
-- `kubernetes/client-go` at 8020fc4 → `/workspace/client-go`
+- `kubernetes/client-go` at v0.32.0 → `/workspace/client-go`
 
 ## Output Format
 
@@ -31,12 +31,14 @@ Create a file at `/workspace/answer.json` with your findings in the following st
 ```json
 {
   "files": [
-    {"repo": "org/repo-name", "path": "relative/path/to/file.go"}
+    {"repo": "sg-benchmarks/kubernetes-client-go", "path": "relative/path/to/file.go"}
   ],
   "text": "Narrative explanation of your findings, citing repos and file paths."
 }
 ```
 
+**Important**: Use `"repo": "sg-benchmarks/kubernetes-client-go"` exactly — this is the canonical repo identifier used by the evaluation oracle. The local checkout at `/workspace/client-go` corresponds to this repo.
+
 Include only the `files` field. Your answer is evaluated against a closed-world oracle — completeness matters.
 
 ## Evaluation

benchmarks/ccb_mcp_crossrepo_tracing/ccx-dep-trace-004/instruction.md

@@ -24,7 +24,7 @@ adding observability, or extending the query pipeline.
 
 The local `/workspace/` directory contains all repositories:
 - `grafana/grafana` at v11.4.0 → `/workspace/grafana`
-- `grafana/loki` at a3af38d → `/workspace/loki`
+- `grafana/loki` at v3.3.4 → `/workspace/loki`
 
 ## Output Format
 
@@ -33,12 +33,18 @@ Create a file at `/workspace/answer.json` with your findings in the following st
 ```json
 {
   "chain": [
-    {"repo": "org/repo-name", "path": "relative/path/to/file.go", "symbol": "TypeOrFunctionName"}
+    {"repo": "grafana/grafana", "path": "relative/path/to/file.go", "symbol": "TypeOrFunctionName"}
   ],
   "text": "Narrative explanation of the call chain, citing specific repos and file paths."
 }
 ```
 
+**Important**: Use exact repo identifiers as they appear in the oracle:
+- For Grafana: `"repo": "grafana/grafana"`
+- For Loki: `"repo": "sg-benchmarks/grafana-loki"`
+
+The local checkout at `/workspace/loki` corresponds to `sg-benchmarks/grafana-loki`.
+
 List the chain steps in order from Grafana (caller) to Loki (callee). Your answer is evaluated
 against a closed-world oracle — precision matters.
 

benchmarks/ccb_mcp_incident/ccx-incident-031/instruction.md

@@ -50,7 +50,7 @@ Create a file at `/workspace/answer.json` with your findings:
 {
   "files": [
     {
-      "repo": "org/repo-name",
+      "repo": "etcd-io/etcd",
       "path": "relative/path/to/file.go",
       "function": "FunctionName"
     }
@@ -59,6 +59,8 @@ Create a file at `/workspace/answer.json` with your findings:
 }
 ```
 
+**Important**: Use `etcd-io/etcd` as the exact `repo` identifier in your answer. The oracle checks for files `server/storage/mvcc/kvstore.go` and `server/storage/mvcc/kvstore_txn.go` in `etcd-io/etcd`. Do not cite vendored copies in `kubernetes/kubernetes`.
+
 ## Evaluation
 
 Your answer will be scored on:

benchmarks/ccb_mcp_onboarding/ccx-explore-042-ds/instruction.md

@@ -41,7 +41,7 @@ Create a file at `/workspace/answer.json` with your findings:
 {
   "chain": [
     {
-      "repo": "org/repo-name",
+      "repo": "numpy/numpy",
       "path": "relative/path/to/file.py",
       "symbol": "FunctionOrClassName",
       "description": "What role this plays in the data flow"
@@ -51,6 +51,8 @@ Create a file at `/workspace/answer.json` with your findings:
 }
 ```
 
+**Important**: Use exact repo identifiers as they appear in Sourcegraph. The oracle expects `repo` values of `numpy/numpy` (array layer), `pandas-dev/pandas` (data structure layer), and `scipy/scipy` (scientific computation layer). The `repo` field must match these exactly.
+
 The `chain` should contain at least 3 steps representing the 3 layers described above.
 
 ## Evaluation

benchmarks/ccb_mcp_onboarding/ccx-onboard-050-ds/instruction.md

@@ -36,7 +36,7 @@ Create a file at `/workspace/answer.json` with your findings:
 {
   "chain": [
     {
-      "repo": "org/repo-name",
+      "repo": "sg-benchmarks/kubernetes-client-go",
       "path": "relative/path/to/file.go",
       "symbol": "FunctionOrTypeName",
       "description": "What this step does in the flow"
@@ -46,6 +46,8 @@ Create a file at `/workspace/answer.json` with your findings:
 }
 ```
 
+**Important**: Use exact repo identifiers as they appear in Sourcegraph. The oracle expects `repo` values of `sg-benchmarks/kubernetes-client-go` (client layer), `kubernetes/kubernetes` (API server layer), and `etcd-io/etcd` (storage layer). The `repo` field must match these exactly.
+
 The `chain` should contain at least 3 steps representing the 3 layers described above.
 
 ## Evaluation

benchmarks/ccb_mcp_platform/ccx-explore-091-ds/instruction.md

@@ -41,7 +41,7 @@ Create a file at `/workspace/answer.json` with your findings:
 {
   "files": [
     {
-      "repo": "org/repo-name",
+      "repo": "sg-benchmarks/kubernetes-api",
       "path": "relative/path/to/file.go",
       "description": "What this file contains and its role in the deployment pattern"
     }
@@ -50,6 +50,8 @@ Create a file at `/workspace/answer.json` with your findings:
 }
 ```
 
+**Important**: Use exact repo identifiers as they appear in Sourcegraph. The oracle expects entries for `sg-benchmarks/kubernetes-api` (API type definitions) and `sg-benchmarks/kubernetes-client-go` (client examples and docs). The `repo` field must match these exactly.
+
 The `files` list should include at least 3 files across 2+ repos that together define
 the canonical service deployment pattern.
 

benchmarks/ccb_mcp_security/ccx-vuln-remed-011/instruction.md

@@ -11,7 +11,7 @@ that declare **`cookie` as a direct runtime dependency** (listed under
 `"dependencies"`, not `"devDependencies"`).
 
 For each match, report:
-- The repository (`org/repo-name`)
+- The repository (e.g., `sg-benchmarks/expressjs-express`)
 - The file path within the repository
 - The version constraint declared for `cookie`
 
@@ -38,7 +38,7 @@ Create a file at `/workspace/answer.json` with your findings:
 {
   "files": [
     {
-      "repo": "org/repo-name",
+      "repo": "sg-benchmarks/expressjs-express",
       "path": "relative/path/to/package.json",
       "version": "the-version-constraint"
     }
@@ -47,6 +47,8 @@ Create a file at `/workspace/answer.json` with your findings:
 }
 ```
 
+**Important**: Use exact repo identifiers as they appear in Sourcegraph. The repos to search are `nodejs/node`, `sg-benchmarks/expressjs-express`, `sg-benchmarks/lodash`, and `sg-benchmarks/prisma-prisma`. Note: the local `/workspace/express` directory maps to `sg-benchmarks/expressjs-express` in Sourcegraph — use `sg-benchmarks/expressjs-express` as the `repo` value in your answer.
+
 Include only entries where `cookie` appears under `"dependencies"` (not `"devDependencies"`
 or `"scripts"`). Your answer is evaluated against a closed-world oracle — completeness matters.
 

benchmarks/ccb_mcp_security/ccx-vuln-remed-014/instruction.md

@@ -12,7 +12,7 @@ and/or Mimir) that register HTTP handlers or routes **without applying
 authentication middleware** such as `middleware.AuthenticateUser`.
 
 For each file found, report:
-- The repository (`org/repo-name`)
+- The repository (e.g., `sg-benchmarks/grafana-loki`)
 - The file path within the repository
 - The specific endpoint path(s) registered without authentication
 
@@ -38,7 +38,7 @@ to surface.
 
 The local `/workspace/` directory contains all repositories:
 - `grafana/grafana` at v11.4.0 → `/workspace/grafana`
-- `grafana/loki` at a3af38d → `/workspace/loki`
+- `grafana/loki` at v3.3.4 → `/workspace/loki`
 
 ## Output Format
 
@@ -48,7 +48,7 @@ Create a file at `/workspace/answer.json` with your findings:
 {
   "files": [
     {
-      "repo": "org/repo-name",
+      "repo": "sg-benchmarks/grafana-loki",
       "path": "relative/path/to/file.go",
       "endpoints": ["/endpoint/path/1", "/endpoint/path/2"]
     }
@@ -57,6 +57,8 @@ Create a file at `/workspace/answer.json` with your findings:
 }
 ```
 
+**Important**: Use exact repo identifiers as they appear in Sourcegraph. The MCP-only repos to audit are `sg-benchmarks/grafana-loki` and `sg-benchmarks/grafana-mimir`. Note: the local `/workspace/loki` directory maps to `sg-benchmarks/grafana-loki` in Sourcegraph — use `sg-benchmarks/grafana-loki` as the `repo` value in your answer. The oracle checks for `pkg/dataobj/explorer/service.go` in `sg-benchmarks/grafana-loki`.
+
 Your answer is evaluated against a closed-world oracle — completeness matters.
 
 ## Evaluation