[Lane A / Phase 0] Cross-pack quality hardening

[ColinLi98]

PR summary

• Lane: Lane A - Cross-pack Quality
• Phase: Phase 0 / Phase 1 quality hardening
• Task: Cross-pack quality diagnostics, long-route readiness, and phase0 guardrail stabilization
• Goal met: yes - extracts Lane A kernel-quality work into its own branch/PR, separate from Agent Studio and .nosbook upload work
• Out-of-scope changes introduced: no

Evidence

• Tests run: GitHub cross-pack-quality run 25756523056 passed; GitHub ops-navigation-stale-ref-smoke run 25756523023 passed; local targeted checks for phase0 guardrails, provider routing/static provider diagnostics, author workflow approval states, learned assisted gate, and longform closeout passed during branch hardening
• Benchmark / eval run: GitHub cross-pack benchmark and cross-pack merge gate passed in run 25756523056
• strongest pack delta: benchmark evidence reports pass-rate stability; strongest/weakest ranking is treated as diagnostic evidence, not a markdown snapshot gate
• weakest pack delta: weakest-pack diagnostics are preserved through benchmark/merge gate; phase0 report comparison no longer fails on runtime ranking noise
• cross-pack pass-rate delta: latest validated cross-pack gate passed; local benchmark evidence during repair showed cross_pack_pass_rate=1.000 and benchmark delta +0.067
• issue category delta (Q03/Q04/Q05/Q09 if relevant): Lane A scope targets Q03 repetition, Q04 over-explanation, Q05 scene detail, and Q09 pacing through content contracts, quality pass/runtime diagnostics, longform benchmark coverage, and guardrail reporting
• rollback point: revert this PR branch or the latest guardrail stabilization commit edfab88 if only the phase0 markdown comparison behavior needs rollback
• next suggested task: review this as Lane A only, then continue with the next Lane A cross-pack/long-route benchmark task after merge

Product impact

• Does this move commercialization forward?: yes
• Does this improve kernel/product/ops instead of just current-pack polish?: yes
• Does this make weakest packs easier to diagnose or improve?: yes

Scope Notes

• This PR is intentionally separate from Agent Studio UX/smoke work and .nosbook upload bridge work.
• Path check found no Agent Studio / nosbook / upload_nosbook / frontend shell smoke files in this branch diff.
• The latest CI failure was fixed by stabilizing phase0 guardrail markdown comparison for runtime-noisy benchmark fields while keeping structural benchmark and merge gate checks intact.