feat: oracle hydration, model fix, protonmail removal, MCP-unique task prep

- Fix critical model default in run_selected_tasks.sh: opus -> haiku
- Hydrate 47 MCP-unique task_spec.json oracles from oracle_answer.json
- Remove 3 protonmail tasks (unresolvable git apply --allow-empty verifier bug)
- Update task.toml timeouts for build/design tasks based on rerun analysis
- Add 4 new MCP-unique tasks (057, 042, 050, 091) with eval.sh verifiers
- Add direct_verifier.sh for 5 ccb_mcp_org tasks (081-083, 122, 127)
- Add batch selection configs: mcp_unique_batch, rerun_haiku_timeout, etc.
- Update MCP task instructions and metadata across all 10 ccb_mcp_* suites
- Add Makefile, analysis scripts, doc updates
- Update generate_manifest.py with MCP-unique suite detection
- Update selected_benchmark_tasks.json and config_utils.py

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

Author: sjarmak

CLAUDE.md

@@ -44,6 +44,7 @@ full operations manual.
 
 ## Maintenance
 - Root and local `AGENTS.md` / `CLAUDE.md` files are generated from sources in `docs/ops/`.
+- `docs/START_HERE_BY_TASK.md` is generated from `docs/ops/task_routes.json`.
 - Regenerate after edits (single command):
 ```bash
 python3 scripts/refresh_agent_navigation.py

Makefile

@@ -0,0 +1,13 @@
+.PHONY: agent-nav agent-nav-check docs-consistency repo-health-quick
+
+agent-nav:
+	python3 scripts/refresh_agent_navigation.py
+
+agent-nav-check:
+	python3 scripts/refresh_agent_navigation.py --check
+
+docs-consistency:
+	python3 scripts/docs_consistency_check.py
+
+repo-health-quick:
+	python3 scripts/repo_health.py --quick

benchmarks/ccb_build/camel-fix-protocol-feat-001/task.toml

@@ -23,7 +23,7 @@ reward_type = "ir_checklist"
 description = "Compilation check + IR metrics (file coverage) + keyword scoring for feature implementation"
 
 [environment]
-build_timeout_sec = 1800.0
+build_timeout_sec = 3600.0
 
 [environment.setup_scripts]
 mcp_config = """#!/bin/bash

benchmarks/ccb_build/flink-pricing-window-feat-001/task.toml

@@ -23,7 +23,7 @@ reward_type = "ir_checklist"
 description = "Compilation check + IR metrics (file coverage) + keyword scoring for feature implementation"
 
 [environment]
-build_timeout_sec = 1800.0
+build_timeout_sec = 3600.0
 
 [environment.setup_scripts]
 mcp_config = """#!/bin/bash

benchmarks/ccb_build/k8s-noschedule-taint-feat-001/task.toml

@@ -21,7 +21,7 @@ command = "bash /tests/test.sh"
 reward_type = "checklist"
 description = "Weighted checklist of required code structure and patterns"
 [environment]
-build_timeout_sec = 1800.0
+build_timeout_sec = 3600.0
 
 [environment.setup_scripts]
 mcp_config = """#!/bin/bash

benchmarks/ccb_build/k8s-score-normalizer-refac-001/task.toml

@@ -23,7 +23,7 @@ reward_type = "ir_checklist"
 description = "IR metrics (file coverage/completeness) + compilation check for cross-file refactoring"
 
 [environment]
-build_timeout_sec = 1800.0
+build_timeout_sec = 3600.0
 
 [environment.setup_scripts]
 mcp_config = """#!/bin/bash

benchmarks/ccb_build/rust-subtype-relation-refac-001/task.toml

@@ -23,7 +23,7 @@ reward_type = "ir_checklist"
 description = "IR metrics (file coverage/completeness) + compilation check for cross-file refactoring"
 
 [environment]
-build_timeout_sec = 1800.0
+build_timeout_sec = 3600.0
 
 [environment.setup_scripts]
 mcp_config = """#!/bin/bash

benchmarks/ccb_build/servo-scrollend-event-feat-001/task.toml

@@ -21,7 +21,7 @@ command = "bash /tests/test.sh"
 reward_type = "checklist"
 description = "Weighted checklist of required code structure and patterns"
 [environment]
-build_timeout_sec = 2400.0
+build_timeout_sec = 3600.0
 
 [environment.setup_scripts]
 mcp_config = """#!/bin/bash

benchmarks/ccb_design/camel-routing-arch-001/task.toml

@@ -23,7 +23,7 @@ reward_type = "ir_checklist"
 description = "IR metrics (file recall/precision) + keyword overlap for architectural analysis"
 
 [environment]
-build_timeout_sec = 1800.0
+build_timeout_sec = 3600.0
 
 [environment.setup_scripts]
 mcp_config = """#!/bin/bash

benchmarks/ccb_design/etcd-grpc-api-upgrade-001/task.toml

@@ -19,7 +19,7 @@ reward_type = "semantic_similarity"
 description = "Semantic similarity of content, file references, and patterns"
 
 [environment]
-build_timeout_sec = 300.0
+build_timeout_sec = 3600.0
 cpus = 2
 memory_mb = 4096
 storage_mb = 10240