T5-P1-A7-WP2 Produce a structured disposition table for the string-to-capability series (#3103–#3137), the four conformance-test issues (#3124, #3273, #3385, #2841), the builder-divergence issue (#3699), and th...

[Trecek]

Goal

Produce a structured disposition table for the string-to-capability series (#3103–#3137), the four conformance-test issues (#3124, #3273, #3385, #2841), the builder-divergence issue (#3699), and the R-G/D9 named subsets — classifying each as SUPERSEDED (closed by delivered commits), ABSORBED (mapped to a specific active program WP with the WP ID noted), or INDEPENDENT (kept open, no program WP owns it). Flag any issue whose absorbing WP has not yet been assigned.

Context

• Phase: Hook Efficacy Probe + Fail-Closed Completions (Milestone: 1-hook-efficacy-probe-fail-closed-completions)
• Assignment: P1-A7 (R-G seed: Disposition pass on top-priority open meta-architecture issues — produce supersession/absorbed/independent classification table)
• Depends on: P1-A7-WP1
• Depended on by: P1-A7-WP3

Deliverables

• Disposition table (embedded in WP output): three-way classification (SUPERSEDED/ABSORBED/INDEPENDENT) for all 35 P1–P6 string-to-capability issues (#3103–#3137), with evidence column (commit SHA or WP ID)
• Disposition entries for conformance-test issues #3124, #3273, #3385, #2841: map against delivered TestCodingAgentBackendConformance (#3991) and flag remaining behavioral coverage gaps as ABSORBED into B2/R-B program requirements
• Disposition entries for builder-divergence issue #3699 and D9 independent bugs (#3297, #3877, #3756, #3876, #3638, #3676, #3836, #3781): ABSORBED or INDEPENDENT with generator tag and notes
• Gap registry: list of ABSORBED issues whose named absorbing program WP has not yet been assigned (unassigned-absorbed gap list, input to P2-A9 WP creation)
• Snapshot annotation: issues created after 2026-06-10 kickoff date identified and tagged for P2-A9 rolling review, not chased in this WP

Technical Steps

1. Read issues_catalog.md and backend-contract-layer-task.md sections R-G and D9 to extract the explicit named issue subsets that are in scope for this WP.
2. For each of the 35 P1–P6 string-to-capability issues ([Backend] P1-A1-WP1: Extend the BackendCapabilities frozen dataclass with 9 new boolean fields (all defaulting to False) so downstream phases have a stable, named capability surface to read. #3103–[Backend] P6-A7-WP1: Add backend.setup_session_dir.return_value = None to the _mock_backend() helper in tests/execution/conftest.py so that all downstream test files using that fixture remain green once P6-A1-WP2 a... #3137), cross-reference the issue description against the live codebase (BackendConventions, AUTOSKILLIT_APPLICABLE_GUARDS, hook_config_format, write_detection_strategy, patch_format, _KNOWN_BACKEND_NAMES, commands.py deprecation shims, etc.) and the recent commit log to determine delivery status.
3. For each delivered item, record the closing commit SHA or PR number and classify as SUPERSEDED.
4. For items not yet delivered, identify which active program WP in P1 (A1–A6) or a later phase owns the requirement, and classify as ABSORBED with the WP ID.
5. For items with no owning program WP, classify as INDEPENDENT and note the gap (unassigned absorbed item).
6. Apply the same three-way disposition to conformance-test issues [Backend] P5-A1-WP1: Add tests/contracts/test_backend_compliance.py containing a single TestBackendCompliance class whose eleven test methods each iterate BACKEND_REGISTRY.values() and assert return-type correctness fo... #3124, Backend-parametrized CLI tests and cross-validation contract coverage #3273, Codex Backend Architectural Immunity — Capability Consumption, Symmetric Env, and NotImplementedError Arch Tests #3385, and Ticket Grouper needs effort-based splitting for multi-file findings #2841: Implementation Plan: Produce the Complete Parametrized Conformance Test Class #3991 (commit 4cadec7) delivered TestCodingAgentBackendConformance with 27 methods and BackendContractBase; map each issue against that delivery plus the remaining behavioral gaps (B2 in R-B) that are not yet covered.
7. Apply the same disposition to builder-divergence issue Rectify: Codex Backend Systemic Builder Divergence — Shared Base, Closed Categorization, and Semantic Flag Validation #3699 against the delivered _cmd_builder.py shared base and remaining P6 consolidation work.
8. For the R-G named D9 independent bugs (Codex config.toml: TOML parse incompatibility causes destructive overwrite of user settings #3297 TOML overwrite, 59% of Codex sessions have expired /dev/shm SKILL.md paths — ephemeral session cleanup race #3877 expired SKILL.md paths, Rectify: Codex NDJSON Parser Schema Drift — Dual-Schema Reconciliation with Observability Layer #3756 NDJSON drift, git worktree add fundamentally broken under Codex workspace-write sandbox — .git/ mounted read-only #3876 .git read-only, Codex tool_timeout_sec: 120s kills all long-running MCP tool calls #3638 MCP tool timeout, CODEX_MODEL_ALIASES Hardcodes Stale Model IDs — Every Codex Session Runs o4-mini Instead of gpt-5.5 #3676 stale model id, run_skill codex_status="not-applicable" misclassification blocks 14 skills from Codex backend #3836/test_check codex_status="not-applicable" misclassification blocks 5 skills from Codex backend #3781 codex_status misclassifications), confirm each is independent (no program WP absorbs it) and flag them as INDEPENDENT with their generator tag.
9. For the R-G named PY1–PY8 and T4 subsets that appear in R-G's explicit list (not the full population), apply the same three-way disposition: map against delivered work or the active P1 WPs that cover the same ground.
10. Assemble a disposition table with columns: issue_number, title_short, disposition (SUPERSEDED|ABSORBED|INDEPENDENT), evidence (commit SHA, PR number, or WP ID), generator, and notes.
11. Flag any ABSORBED entry whose named WP does not yet appear in the program plan — these are unassigned absorbed gaps that need WP creation in a later phase.
12. Note the snapshot rule: this disposition covers issues open at program kickoff (2026-06-10); issues created after that date are tagged for P2-A9 review.

Acceptance Criteria

• Every one of the 35 P1–P6 string-to-capability issues ([Backend] P1-A1-WP1: Extend the BackendCapabilities frozen dataclass with 9 new boolean fields (all defaulting to False) so downstream phases have a stable, named capability surface to read. #3103–[Backend] P6-A7-WP1: Add backend.setup_session_dir.return_value = None to the _mock_backend() helper in tests/execution/conftest.py so that all downstream test files using that fixture remain green once P6-A1-WP2 a... #3137) carries a disposition of SUPERSEDED, ABSORBED, or INDEPENDENT with a non-empty evidence field (commit SHA, PR number, or WP ID).
• Issues [Backend] P5-A1-WP1: Add tests/contracts/test_backend_compliance.py containing a single TestBackendCompliance class whose eleven test methods each iterate BACKEND_REGISTRY.values() and assert return-type correctness fo... #3124, Backend-parametrized CLI tests and cross-validation contract coverage #3273, Codex Backend Architectural Immunity — Capability Consumption, Symmetric Env, and NotImplementedError Arch Tests #3385, and Ticket Grouper needs effort-based splitting for multi-file findings #2841 each carry a disposition with explicit reference to commit 4cadec7 (TestCodingAgentBackendConformance, Implementation Plan: Produce the Complete Parametrized Conformance Test Class #3991) and/or the specific behavioral gap (B2) that remains uncovered.
• Issue Rectify: Codex Backend Systemic Builder Divergence — Shared Base, Closed Categorization, and Semantic Flag Validation #3699 (builder-divergence) carries a disposition referencing either the delivered _cmd_builder.py shared base or a specific active program WP for remaining consolidation work.
• All eight D9 independent bugs (Codex config.toml: TOML parse incompatibility causes destructive overwrite of user settings #3297, 59% of Codex sessions have expired /dev/shm SKILL.md paths — ephemeral session cleanup race #3877, Rectify: Codex NDJSON Parser Schema Drift — Dual-Schema Reconciliation with Observability Layer #3756, git worktree add fundamentally broken under Codex workspace-write sandbox — .git/ mounted read-only #3876, Codex tool_timeout_sec: 120s kills all long-running MCP tool calls #3638, CODEX_MODEL_ALIASES Hardcodes Stale Model IDs — Every Codex Session Runs o4-mini Instead of gpt-5.5 #3676, run_skill codex_status="not-applicable" misclassification blocks 14 skills from Codex backend #3836, test_check codex_status="not-applicable" misclassification blocks 5 skills from Codex backend #3781) are classified INDEPENDENT with no active program WP absorbing them, consistent with the task file's characterization of them as 'independent open bugs that survive the rework'.
• Any ABSORBED issue whose absorbing program WP ID does not appear in the P1 sibling WP list is explicitly flagged as an unassigned-absorbed gap in the gap registry.
• The disposition table covers only issues open at the 2026-06-10 kickoff snapshot; issues created after that date are not individually dispositioned but are noted for P2-A9.
• No exhaustive triage of the full PY1–PY8 (PX1 P1-A1-WP1: Delete the two list entries appending CodexFlags #2667–75% of zero_writes classifications are false positives since CC 2.1.140 — systematic parse_session_result runtime failure #2714) or T4 (T4-P1-A1-WP1 Seal CodexEventType with an ITEM_UPDATED member and route it to silent discard in both _scan_codex_ndjson and CodexStreamParser.parse_line, preventing item.updated NDJSON lines from reaching the UNKNOWN warning path. #3916–T4-P4-A5-WP1 Add a regression test file covering five gap areas not exercised by test_cmd_builder.py: frozen-instance enforcement, cmd-element type contract (all str), CmdOrigin field-type shapes (binary=str, outer containers=tuple), direct-construction origin-None default, and empty-binary ValueError. #3937) populations is performed; only the R-G and D9 explicitly named subsets are covered.