P6.0-P6.2 runtime benchmark contract scaffold

[Harzva]

Summary

• Adds the P6.0 benchmark taxonomy contract for mobile task categories, runtime assertions, and score dimensions.
• Adds the P6.1 communication substrate contract: MailboxMessage, EvidenceLedgerEntry, RuntimeEvent, MemoryCommitProposal, and role allowlists/budgets/input filters/return contracts.
• Adds the P6.2 runtime verifier scaffold and a validator-compatible non-counted run package.

Non-counted Boundary

This is a P6.0-P6.2 contract/scaffold upgrade only. It is not a formal benchmark and must not be reported as counted strategy ablation evidence.

• run_kind=strategy_pilot_not_counted
• counts_as_experiment=false
• counts_as_strategy_ablation_result=false
• no model callback, provider claim, device action, raw transcript, or secret material is included

Evidence

• Contract doc: docs/mobile-harness-benchmark/strategy-ablation/p60-p62-runtime-benchmark-contract.md
• Run package: docs/mobile-harness-benchmark/strategy-ablation/runs/p60-p62-runtime-benchmark-upgrade/
• Run JSON: docs/mobile-harness-benchmark/strategy-ablation/runs/p60-p62-runtime-benchmark-upgrade/run.json
• Verifier JSON: docs/mobile-harness-benchmark/strategy-ablation/runs/p60-p62-runtime-benchmark-upgrade/runtime_benchmark_upgrade_verifier.json
• Summary: docs/mobile-harness-benchmark/strategy-ablation/runs/p60-p62-runtime-benchmark-upgrade/summary.md

Verification

• python3 -m py_compile scripts/run_mobile_harness_strategy_p60_p62_runtime_benchmark_upgrade.py scripts/validate_mobile_harness_strategy_ablation.py
• python3 scripts/validate_mobile_harness_strategy_ablation.py --registry docs/mobile-harness-benchmark/strategy-ablation/strategy_registry.json --run docs/mobile-harness-benchmark/strategy-ablation/runs/p60-p62-runtime-benchmark-upgrade/run.json
  • PASS, 6 strategies, 18 results, strategy_pilot_not_counted
• git diff HEAD^..HEAD --check
• staged/commit secret and private path scan: no findings
• boundary assertions passed for counts_as_experiment=false and counts_as_strategy_ablation_result=false

Review Focus

• scripts/run_mobile_harness_strategy_p60_p62_runtime_benchmark_upgrade.py
• docs/mobile-harness-benchmark/strategy-ablation/p60-p62-runtime-benchmark-contract.md
• docs/mobile-harness-benchmark/strategy-ablation/runs/p60-p62-runtime-benchmark-upgrade/
• docs/mobile-harness-roadmp/roadmp2026.6.20-p6-mobile-agent-coevolution.md
• docs/mobile-harness-roadmp/roadmp2026.6.20.md

Next Step After Merge

Enter P6.3 Android real device lane: install the latest APK, grant or verify MobileCode Phone Use Accessibility service, run dry/action probes, and save screenshot, UI XML, logcat, focus state, and WebView state assertions. Keep that lane non-counted until repeated samples and promotion gates pass.