Track backlog simplification and Ruby maintenance priorities #822

Companion tracker: agent-coordination #315.

This maintainer-requested tracker preserves the backlog complexity audit and the recommended order of work. Existing issues and PRs remain the implementation records; this issue consolidates decisions and progress.

The audit covered every open issue and PR in the September 9 snapshot, including issues without PRs. Labels were applied and read back successfully. The audit is a prioritization assessment, not a current-head merge approval. Status updates below were checked September 10, 2026; recheck linked items before acting.

How to use the labels

• complexity:simplify: removes or consolidates ongoing code, concepts, obligations, or operational work.
• complexity:neutral: preserves roughly the same maintenance surface.
• complexity:complexify: adds a maintenance surface. A necessary correctness repair can still belong here; this is not an automatic rejection.
• P1 / P2 / P3: relative priority in the audited backlog, not permission to merge.
• triage:drain-first: bounded existing work worth finishing; triage:simplify-next: reserve cleanup capacity; triage:needs-scope: narrow before implementation; triage:reconcile: compare with shipped work or competing approaches; triage:park: defer pending concrete evidence of value.

Optimize for fewer maintainer interventions and less repeated work, rather than the number of PRs merged. File overlap warrants integration review; it does not automatically establish a semantic dependency. Preserve existing approvals and ownership when reconciling the backlog.

Completed

• Audit and label 184 items: 58 PRs and 126 issues, including their relationships and issues without PRs.
• Classify 39 as simplify, 48 as neutral, and 97 as complexify.
• Separate integration checks from full release verification #808 merged: separate integration checks from full release verification.
• Make ordinary PR closeout lightweight; reserve additional audits for explicit need #816 completed through merged Make ordinary PR closeout lightweight #818: make ordinary PR closeout lightweight.

Recommended order and remaining decisions

1. Cap the drain phase. Finish bounded closeout for Skip Claude reviews for draft pull requests #804 and the remaining Limit optional review nits to the initial fix pass #811 work after merged Separate integration checks from full release verification #808. Reconcile pr-batch: do not chase reviewer re-review prompts when the merge gate is already satisfied #708 and Prevent QA lanes from duplicating targets #706 / Clarify terminal closeout target forms #703 where review, QA, and target instructions disagree. This is not a requirement to clear every open PR before simplification.
2. Retain concrete safety repairs without creating universal gates. Enforce actor trust at the review-data ingestion boundary #799 has trust-boundary and deduplication value. Preserve the needed behavior in Follow-up: calibration GitHub reader needs UTF-8 and surrogate gates #700 and Fail closed on malformed trust role values #717; Enforce actor trust at the review-data ingestion boundary #799 removes an insertion point used by Fail closed on malformed trust role values #717, so adapt the surviving implementation. Fix current-integration identity and live-base policy provenance #814 can preempt the queue if it addresses demonstrated active false-stale blockage, but is not a prerequisite for everything else.
3. Make retirement decisions now. Recommend stopping rebases and not merging Fail closed on unsettled configured reviews #469 / Distinguish an unavailable configured-review gate from unsettled reviews #665 as-is. Preserve minimal acceptance cases, the real Follow-up: Fail closed on unsettled or failed configured reviews before merge #249 defect, and a replacement owner/scope before retiring the unwanted architecture. Do not require equivalence to unwanted architecture. For Fix issue-targeted batch publication #643, compare only the residual behavior after closed Follow-up: completed-batch-publication-preflight cannot publish an issue-targeted lane — three independent defects #522 and shipped Fix issue-targeted completed-batch publication #439; full equivalence has not been established. Reconcile Follow-up: ready-to-draft toggle collapses same-head Validate runs and drops the ready run's evidence #790 against merged Prevent false-ready CI verdicts after draft transitions #792, Block foreign-target lane takeover through cross-task delegation #403 against Add fail-closed target membership guard #425, and the contract portion of Machine-enforce OC-v1: contract test and a non-self-counted narration-volume helper #480 against Add OC-v1 contract coverage #705. Remaining scope must be evaluated separately.
4. Reserve skill simplification capacity. With Make ordinary PR closeout lightweight; reserve additional audits for explicit need #816 / Make ordinary PR closeout lightweight #818 shipped, continue Gate follow-up issue creation with the evaluate-issue rubric and add a sweep that closes low-value agent-filed follow-ups #781 and verify adoption of the lighter closeout behavior. Decide which obligations survive before building more receipts, writers, or validators around them. Narrow Make PR-batch launch prompts human-readable #575 to the useful short-prompt need in Slim the goal-prompt template: pin workflow contracts instead of restating them, and make density host-budget aware #476; avoid adding a ledger, digest, and exact-prose checker merely to support it.
5. Refactor Ruby in bounded, settled areas under Refactor oversized Ruby files and re-enable default RuboCop metrics #309. Follow-up: extract one bounded process-group lifecycle instead of three copies #659 can consolidate process-group handling across three existing callers after the surviving Harden fail-closed workflow replay controls #288 / Bound replay lifecycles and report cleanup honestly #655 semantics settle. Extract parser phases in Follow-up: Extract phases from closeout-evidence-replay HTTPS label parser #631, then remove open-ended word lists in Follow-up: HTTPS prose-label acceptance is an open-ended wording list #632. Revisit Follow-up: pass the trusted applicability proof and its digest as one value object #645 after deciding which receipt values are retained. upgrade-agent-workflows snapshots the whole agent home with rsync; unusable on large ~/.codex #735 can remove full-home copying in favor of a managed-path snapshot for the observed upgrade failure, after installer overlap review. Add a RuboCop metrics baseline ratchet #692 is a lint baseline, not completion of the Ruby refactor.
6. Park expansion until it earns its maintenance cost. Broad budget/task-control work in Enforce hierarchical token budgets for PR batches #426 / Default PR batches to one canonical target #458, release lifecycle machinery in Add a verified stable release channel #809, the custom JavaScript parser in Add non-failing linter settings advice #696, prompt interpretation in Add runtime prompt compatibility gate #697, and broader launcher/capacity/provenance/telemetry machinery should not displace the cleanup slot. Retain the bounded approved Accept trusted-base provenance for merged changes #523 repair without reviving the whole historical program.

Progress checklist

• Resolve the capped drain set and record any residual blockers on the owning items.
• Record explicit keep/narrow/retire decisions for the overlapping large PRs.
• Complete the retained Gate follow-up issue creation with the evaluate-issue rubric and add a sweep that closes low-value agent-filed follow-ups #781 skill cleanup and remove obsolete obligations rather than relocating them.
• Deliver bounded Refactor oversized Ruby files and re-enable default RuboCop metrics #309 refactors with behavior preserved and duplicate code removed.
• Revisit parked work only when an observed need justifies its added maintenance cost.

#809 is not a prerequisite for #808, and #816 was not a prerequisite for #811. Cross-references and ordering suggestions from the audit must not be converted wholesale into an executable dependency graph.

Scope and evidence limits

The original source snapshot contained 89 production Ruby files / 45,015 lines and 36 SKILL.md files / 6,697 lines; 12 production Ruby files exceeded 1,000 lines. These historical counts include comments and blank lines and precede subsequent merges. They identify review targets, not a quality score. Stacked PR diffs must not be summed as independent growth. The substantive objective is fewer concepts, duplicated rules, parsing paths, and maintenance obligations.

Complete audited inventory

This table preserves the original audit classifications. It is a historical snapshot, not a statement that every item remains open. The progress section above incorporates later verified merges; linked items and current labels govern subsequent work.

Completion condition

Close this tracker when the bounded drain and cleanup decisions above have durable outcomes on their owning items. Completion does not require merging every audited PR or implementing parked proposals.

September 10 follow-up: approved fixture repair

• Prevent temporary-pack hardlink races in stack fixture clones #823 is the canonical, approved repair for temporary-pack hardlink races in the two stack fixture clone constructors. Prioritize it as P1, complexity:neutral, triage:drain-first: it reduces repeated CI work with a small change to existing fixture behavior.
• Keep this repair separate from Follow-up: Fail closed on unsettled or failed configured reviews before merge #249 and the existing Recognize freshly verified already-integrated PR heads in current-integration evidence #813 / Fix current-integration identity and live-base policy provenance #814 scope. The existing coordinator owns implementation assignment; this tracker creates no additional lane.
• Controlled reproduction supports the mechanism and committed identity preservation. The actual hosted writer and full-suite cost remain unknown; require hosted evidence before marking that failure resolved.