[realmarcin]

Summary

Refreshes NEXT_TASKS.md after PR #912 (922fae2) using the existing reconcile-next-tasks skill.

• Snapshot 2026-10-18 07:04 UTC: main = 922fae2; 8 open issues, 0 open PRs, list lengths checked against REST search and GraphQL totals.
• PR Correct quality-reporting guidance and advance the work plan #912's closures of Correct predicted-confidence interpretation and pLDDT conversion in the quality guide #907, Remove unsupported quality and refinement verdicts from the reporting guide #908, Correct quality-guide tool independence, literature and 1SAR coverage claims #909, Align quality-reporting instructions with schema and emitter capabilities #910 and Correct residual QDS summary-routing and ensemble instructions #913 were verified on the exact merged PR. The guide correction moves from Tracked on GitHub to Delivered prerequisites, scoped as documentation only (its schema/model diff is description-only; no runtime behavior, adopted threshold or frozen EVAL/QDS record changed).
• The former first recommended direction is removed as delivered; the before/after assessment and tools-and-measurements guide remain proposed / no issue.
• Audit and reconcile ProSMART Ramachandran-Z recommendation and assumptions #911, P2: native Reduce status 1 conflates zero-change completion with abandoned optimization #876, P2: shared-H study must retain native Reduce status masked by PHENIX dispatcher #873, P2: proposed shared-H study cannot bound detached PHENIX helper groups #872, Standalone reduce in the benchmarks most likely ran with no het dictionary; records attribute the effect to 'different dictionaries' #799, Opt-in: decompose the matched-convention clashscore residual (H construction vs dictionaries vs counting) #790, Opt-in: rerun the T15 secondary-structure agreement benchmark with exact per-assigner denominators #765 and [blocked] P3a / round 43: cross-version PHENIX reproducibility round needs a second PHENIX build #760 keep their dispositions; none had a new comment since the 2026-10-17 reconciliation.

No GitHub issue state, scientific record, threshold or registry entry is changed.

Issue references

• Intended closures: none.
• Remaining work (stays open): Audit and reconcile ProSMART Ramachandran-Z recommendation and assumptions #911, P2: native Reduce status 1 conflates zero-change completion with abandoned optimization #876, P2: shared-H study must retain native Reduce status masked by PHENIX dispatcher #873, P2: proposed shared-H study cannot bound detached PHENIX helper groups #872, Standalone reduce in the benchmarks most likely ran with no het dictionary; records attribute the effect to 'different dictionaries' #799, Opt-in: decompose the matched-convention clashscore residual (H construction vs dictionaries vs counting) #790, Opt-in: rerun the T15 secondary-structure agreement benchmark with exact per-assigner denominators #765, [blocked] P3a / round 43: cross-version PHENIX reproducibility round needs a second PHENIX build #760.

Validation

• uv run --locked --extra benchmark -- bash scripts/validate.sh — exit 0 (Python 3.12).
• check_next_tasks_dates.py, check_summary_coverage.py, check_threshold_evidence_status.py, git diff --check — pass.

🤖 Generated with Claude Code