fix(review): stop tool-free cross-model reviews when Claude invokes advisor

[BaseInfinity]

Problem

The Opus 4.8 xhigh fallback reviewer twice violated the intended tool-free review boundary. The runner launched Claude with --safe-mode --tools "" --disable-slash-commands --strict-mcp-config, but Claude still emitted a server_tool_use event for the built-in advisor tool.

Both attempts then consumed the full 600-second wall timeout instead of returning the required structured code-review verdict. The second attempt spent roughly 39k thinking tokens before requesting advisor.

This is reviewer infrastructure failure, not a finding on the frozen #109 candidate. The candidate did not change, its canonical proof remained 11/11, and no review receipt was created.

Related: #109, #137, #143.

Evidence

• Reproduced twice on the same frozen staged tree.
• Sol High completed normally.
• Claude stream emitted server_tool_use with name: advisor despite the declared tool-free CLI flags.
• Runner correctly ended TIMED_OUT, exit 2, with no certification.
• Current log: .git/worktrees/codex-sdlc-issue109-v4/codex-sdlc/review-logs/1786937693307-915130665ff2/cross-model-initial-attempt-1.log.
• Local claude --help documents --tools "" as disabling all tools; official CLI docs also expose allow/deny tool controls.

Acceptance criteria

• Add a focused RED regression that feeds a Claude stream containing server_tool_use: advisor.
• A tool-free cross-model reviewer must reject any tool-use event immediately; it must not wait for the wall timeout.
• Classify the attempt truthfully as reviewer infrastructure/invalid execution, never CLEAN/CERTIFIED.
• Terminate and clean the full reviewer process tree.
• Retain the structured failure summary and bound log.
• Do not invalidate or rerun an unchanged candidate's ordinary proof solely because reviewer infrastructure failed.
• Keep the Opus 4.8 xhigh subscription fallback available; do not silently downgrade or fabricate a receipt.
• Focused tests and canonical proof pass before merge.

Non-goals

• No new review engine.
• No third automatic reviewer attempt.
• No change to Implement incremental corrective commits with a final whole-PR review gate #109's candidate in this issue.

[BaseInfinity]

Implementation status: RED/GREEN complete, adapter integration 115/115, canonical proof 11/11. The live completion gate did not exercise the advisor rejection because Fable first returned a quota-exhausted stream-json envelope with API status 429 but process exit 0. The runner mislabeled that as INVALID_VERDICT and did not enter the approved Opus fallback. #158 remains frozen and unchanged while the pre-existing #143 fallback contract is repaired.

[BaseInfinity]

Tracker checkpoint after the post-#159 completion gate (2026-08-16 23:29 PDT): candidate tree f43b24de8058b9a97e8ac009b7718ccb66361b3b remains staged and unchanged. Focused proof and the canonical 11/11 proof were green. The joint Sol High + Fable High gate returned NOT CERTIFIED. Blocking P1s: the detector inspects top-level fields instead of real nested message.content[] tool-use blocks, and a rejected event can still be overwritten by a parsed clean result. The gate also identified rejection-versus-timeout precedence as a bounded P2. Next action is one corrective TDD slice in /private/tmp/codex-sdlc-issue158, then focused proof, one fresh canonical proof, and one bounded completion gate. No commit, push, or PR exists for this candidate.