Add public claim readiness gates

[ictechgy]

Summary

• Add public_claim_readiness report/dashboard gates for public hosted savings claims.
• Require matched successful tasks, provider-measured primary tokens/cost, quality non-inferiority, shifted-cost accounting, explicit confidence/failure notes, and complete provider-export provenance before claim_allowed=true.
• Keep synthetic/manual/CSV-only/proxy-only evidence claim-safe and update docs/examples/tests.

Validation

• python3 scripts/sync_plugin_copies.py --check
• python3 -m py_compile context-guard-kit/benchmark_runner.py plugins/context-guard/bin/context-guard-bench tests/test_context_guard_kit.py scripts/release_smoke.py
• PYTHONDONTWRITEBYTECODE=1 python3 -m unittest -k benchmark tests.test_context_guard_kit.BenchmarkRunnerTests (35 tests)
• 12-task evidence replay smoke verified public_claim_readiness JSON/dashboard and blocked synthetic replay claims
• python3 scripts/release_smoke.py --timeout 20
• PYTHONDONTWRITEBYTECODE=1 python3 scripts/prepublish_check.py (697 tests)
• git diff --check

Ralplan evidence

• Plan: .omx/plans/ralplan-g006-public-savings-claim-evidence-gates.md
• Architect: .omx/artifacts/ralplan-g006-architect-20260615T081040Z.md (APPROVE/no blockers)
• Critic: .omx/artifacts/ralplan-g006-critic-20260615T081409Z.md (APPROVE/no blockers)

[ictechgy]

G006 quad review + validation evidence

Objective: add fail-closed public_claim_readiness gates so public hosted token/cost savings claims are only allowed when public_claim_readiness.claim_allowed=true after matched successful tasks, provider-measured token/cost, quality non-inferiority, shifted-cost accounting, confidence/failure notes, and provider-export provenance gates all pass.

Validation:

• python3 scripts/sync_plugin_copies.py --check ✅
• python3 -m py_compile context-guard-kit/benchmark_runner.py plugins/context-guard/bin/context-guard-bench tests/test_context_guard_kit.py scripts/release_smoke.py ✅
• PYTHONDONTWRITEBYTECODE=1 python3 -m unittest -k benchmark tests.test_context_guard_kit.BenchmarkRunnerTests ✅ 35 tests
• 12-task evidence replay smoke ✅ synthetic replay claims blocked with provider_measured_token_cost, shifted_cost_accounting, provider_export_provenance
• python3 scripts/release_smoke.py --timeout 20 ✅
• PYTHONDONTWRITEBYTECODE=1 python3 scripts/prepublish_check.py ✅ 697 tests
• git diff --check ✅
• PR CI run 27536622390 ✅: 3.11, 3.12, macOS 3.12 all passed

Quad review loop:

• Codex: APPROVE, no blockers (.omx/artifacts/quad-review-pr207-codex-20260615T093223Z.md)
• Claude: APPROVE, no blockers (.omx/artifacts/claude-you-are-reviewing-pr-207-for-contextguard-task-g006-objectiv-2026-06-15T09-41-35-068Z.md)
• Forge: APPROVE, no blockers (.omx/artifacts/quad-review-pr207-forge-20260615T093223Z.md)
• Agy: APPROVE, no blockers (.omx/artifacts/quad-review-pr207-agy-fallback-20260615T094458Z.md)

Nonblocking follow-ups captured: optionally surface claim_allowed in the dashboard top summary; optionally add extra negative release-smoke/test coverage. No merge blockers.