This is a public-safe derivative of a frozen openai/codex backlog audit. Its primary purpose is to save maintainer triage time by making the full backlog searchable and grouped. It contains no private repository source and no copied raw issue-body corpus. Every row links back to the public upstream GitHub issue.
Snapshot: openai/codex; 11,813 open issues collected from 2026-08-09T00:49:53.987670+00:00 through 2026-08-09T00:53:30.675560+00:00. This is a frozen snapshot: later comments, closures, and newly opened issues are not reflected until a new release is published.
The index covers 31 technical domains, 161 surface cohorts, and 1,081 taxonomy groups.
- ISSUE_INDEX.csv — one row for every frozen issue, including activity/date fields, classification, grouping, evidence depth, and suggested workflow.
- TAXONOMY_INDEX.csv — grouped backlog view for choosing a technical work area.
- TRACKER_INDEX.csv — representative/evidence-rich issues for engineering, product, and docs groups.
- SOLUTION_INDEX.csv — 11 issues with concrete resolution/fix evidence or an explicit fix-verification gate.
- PILOT_GROUPS.md — five concrete groups for a quick maintainer spot-check.
- DATA_DICTIONARY.md — definitions for every master-index column.
- SNAPSHOT.json and PROVENANCE.md — exact collection boundary and source provenance.
- EVIDENCE_INDEX.csv and evidence/ — 14 deep technical evidence packets.
- RETAINED_OPEN_SET.csv and CONSOLIDATION_QUEUE.csv — optional cleanup-model views; secondary to the index.
- METHODOLOGY.md — interpretation and safety boundaries.
- CHECKSUMS.sha256 — integrity hashes for the published bundle.
For an individual issue, ISSUE_INDEX.csv provides the upstream issue URL/title, created/updated timestamps, days since update, comment/reaction counts, labels, technical domain, surface cohort, symptom class, taxonomy, family placement, mapping confidence/basis, review/evidence depth, audit disposition, duplicate authority where proven, deep-evidence references, recommended workflow, and optional cleanup recommendation.
For a work area, TAXONOMY_INDEX.csv shows the grouped backlog so a maintainer can start from related reports instead of searching the issue tracker from scratch.
SOLUTION_INDEX.csv is intentionally small: it contains only 11 issues where the audit captured resolution/fix evidence or a concrete verification gate. Blank solution fields in ISSUE_INDEX.csv mean no defensible solution was established, not that no solution exists.
The bundle does not claim every issue received the same depth of manual root-cause analysis. review_depth explicitly separates manually promoted, curated, deep-packet-referenced, and conservative report-evidence-gate rows.
100% indexed, mapped, and routed does not mean 100% root-caused. Taxonomy/family similarity never grants duplicate authority by itself.
PILOT_GROUPS.md contains five high-volume groups selected for a small maintainer spot-check. No bulk cleanup is required to evaluate the contribution.
A separate scenario models large backlog reduction through evidence-backed closures, bounded information requests, and policy-gated administrative consolidation. It is intentionally secondary to the indexing contribution. Administrative consolidation is not a technical duplicate/fix claim and requires maintainer approval before use.