Implementation Plan: T5-P1-A7-WP2 Disposition Table — Conformance Evidence Gap Fix (#4135)

## Summary

The WP2 disposition table
(`.autoskillit/temp/p1_a7_wp2_disposition_table.md`) was delivered in PR
#4132 and covers 49 unique issues across the string-to-capability
series, conformance-test issues, builder-divergence, guard-bypass, and
D9 independent bugs. Acceptance criteria audit reveals one gap:
**Criterion 2** requires the four conformance-test issues (#3124, #3273,
#3385, #2841) to carry explicit references to commit `4cadec7f8`
(TestCodingAgentBackendConformance / #3991) and/or the specific B2
behavioral gap that remains uncovered. Currently, only #3124 satisfies
this — #3273, #3385, and #2841 are missing the required
cross-references.

The fix updates three rows in §3 of the existing disposition table to
add the mandated `4cadec7f8` and B2 references. No code changes are
required.

## Implementation Plan

Plan file:
`/home/talon/projects/autoskillit-runs/impl-20260628-001243-698616/.autoskillit/temp/make-plan/t5_p1_a7_wp2_disposition_table_plan_2026-06-28_012530.md`

Closes #4005

🤖 Generated with [Claude Code](https://claude.com/claude-code) via
AutoSkillit
<!-- autoskillit:pipeline-signature
steps=prepare_pr,run_arch_lenses,compose_pr,annotate_pr_diff,review_pr
-->

## Token Usage Summary

| Step | Model | count | uncached | output | cache_read | peak_ctx |
turns | cache_write | time |

|------|-------|-------|----------|--------|------------|----------|-------|-------------|------|
| plan* | opus[1m] | 1 | 7.1k | 8.8k | 915.6k | 76.2k | 32 | 60.7k | 7m
26s |
| verify* | sonnet | 1 | 52 | 6.6k | 210.6k | 46.7k | 17 | 28.3k | 2m
51s |
| implement* | MiniMax-M3 | 1 | 49.7k | 4.7k | 606.5k | 0 | 38 | 0 | 1m
10s |
| merge_gate_fix* | sonnet | 1 | 110 | 5.5k | 527.3k | 47.6k | 35 |
29.5k | 2m 55s |
| audit_impl* | sonnet | 1 | 380 | 4.2k | 199.3k | 40.1k | 13 | 25.5k |
3m 46s |
| prepare_pr* | MiniMax-M3 | 1 | 41.7k | 2.1k | 188.5k | 0 | 13 | 0 |
52s |
| compose_pr* | MiniMax-M3 | 1 | 37.9k | 1.4k | 283.7k | 0 | 13 | 0 |
55s |
| **Total** | | | 136.9k | 33.3k | 2.9M | 76.2k | | 144.0k | 19m 57s |

\* *Step used a non-Anthropic provider; caching behavior may differ.*

## Token Efficiency

| Step | LoC Changed | cache_read/LoC | cache_write/LoC | output/LoC |
|------|-------------|----------------|-----------------|------------|
| plan | 0 | — | — | — |
| verify | 0 | — | — | — |
| implement | 7 | 86649.3 | 0.0 | 667.1 |
| merge_gate_fix | 7 | 75335.4 | 4210.6 | 791.0 |
| audit_impl | 0 | — | — | — |
| prepare_pr | 0 | — | — | — |
| compose_pr | 0 | — | — | — |
| **Total** | **14** | 209405.4 | 10287.1 | 2380.7 |

## Model Usage Breakdown

| Model | steps | uncached | output | cache_read | cache_write | time |
|-------|-------|----------|--------|------------|-------------|------|
| opus[1m] | 1 | 7.1k | 8.8k | 915.6k | 60.7k | 7m 26s |
| sonnet | 3 | 542 | 16.3k | 937.2k | 83.3k | 9m 33s |
| MiniMax-M3 | 3 | 129.2k | 8.2k | 1.1M | 0 | 2m 57s |

Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: Trecek

.autoskillit/temp/p1_a7_wp2_disposition_table.md

@@ -55,9 +55,9 @@ Columns: `issue_number | title_short | disposition | evidence | generator | note
 | Issue | Title Short | Disposition | Evidence | Generator | Notes |
 |-------|-------------|-------------|----------|-----------|-------|
 | #3124 | test_backend_compliance.py | SUPERSEDED | `95b745031` (PR #3594); superseded by `4cadec7f8` (PR #3991) which delivered the full conformance class | R-G/P5 | Original compliance file evolved into `TestCodingAgentBackendConformance` with 31 methods (extended by PR #4081). **Counted once in §2 above.** |
-| #3273 | Backend-parametrized CLI tests | SUPERSEDED | `4537136a8` (PR #3299); conformance class parametrized over `BACKEND_REGISTRY` covers both backends | R-G | Cross-validation contract coverage delivered by parametrized conformance tests |
-| #3385 | Codex Backend Architectural Immunity | ABSORBED | WP1 mapped to R-A/A1, R-A/A3; 8 forward-declared capability fields remain without production consumers; `CodexBackend.build_inspector_cmd()` at `codex.py:1117-1119` raises `AssertionError("inspector_capable is True but build_inspector_cmd has no implementation")` | R-G | Partially delivered: capability consumption arch test exists (`test_capability_consumption.py`) but enforcement completeness gaps remain |
-| #2841 | Ticket Grouper effort-based splitting | INDEPENDENT | N/A | R-G/PY8 | **Discrepancy:** PY8 referenced this as "Conformance test class parametrized over both backends" but current issue content is about audit skill's ticket grouper (effort-based splitting for multi-file findings). Current title/content is INDEPENDENT — no program WP covers ticket grouper improvements. If intended as conformance: SUPERSEDED by `4cadec7f8` |
+| #3273 | Backend-parametrized CLI tests | SUPERSEDED | `4537136a8` (PR #3299); `4cadec7f8` (PR #3991) delivered full `TestCodingAgentBackendConformance` with 33 methods parametrized over `BACKEND_REGISTRY`; extended by PR #4081 (B2 behavioral conformance suite) | R-G | Cross-validation contract coverage delivered by parametrized conformance tests. B2 remaining gap: env injection presence, resume-id inclusion, and flag semantics behavioral tests not yet exhaustive per R-B/B2 spec |
+| #3385 | Codex Backend Architectural Immunity | ABSORBED | WP1 mapped to R-A/A1, R-A/A3; `4cadec7f8` (PR #3991) delivered `TestCodingAgentBackendConformance` with structural conformance coverage; 8 forward-declared capability fields remain without production consumers; `CodexBackend.build_inspector_cmd()` at `codex.py:1117-1119` raises `AssertionError` | R-G | Partially delivered: conformance class (`4cadec7f8`) covers type-level contracts. B2 remaining gap: behavioral conformance (env injection presence, symmetric env forwarding, NotImplementedError arch tests for capability-gated methods) not yet covered per R-B/B2 spec. Enforcement completeness gaps remain under R-A/A1 and R-A/A3 |
+| #2841 | Ticket Grouper effort-based splitting | INDEPENDENT | N/A | R-G/PY8 | **Discrepancy:** PY8 referenced this as "Conformance test class parametrized over both backends" but current issue content is about audit skill's ticket grouper (effort-based splitting for multi-file findings). Current title/content is INDEPENDENT — no program WP covers ticket grouper improvements. If intended as conformance: SUPERSEDED by `4cadec7f8` (PR #3991, `TestCodingAgentBackendConformance` with 33 methods). B2 note: conformance behavioral gaps (R-B/B2) are tracked separately and do not affect this issue's INDEPENDENT classification |
 
 ### §4 — Builder-Divergence (#3699)
 
@@ -132,6 +132,7 @@ To identify post-kickoff issues in scope clusters, run:
 3. **#2841 issue check:** Current title is "Ticket Grouper needs effort-based splitting for multi-file findings". Body describes audit skill Ticket Grouper improvements, not conformance tests. Classified INDEPENDENT per current content.
 4. **#3699 codebase check:** `BackendCmdBuilderBase` ABC confirmed at `src/autoskillit/execution/backends/_backend_cmd_builder_base.py:64`. `CodexBackend` and `ClaudeCodeBackend` both inherit from it. Notes updated to reflect further consolidation progress.
 5. **#3124 double-listing:** Appears in both §2 (string-to-capability) and §3 (conformance) but counted once in summary statistics.
+6. **Criterion 2 conformance cross-references (added 2026-06-28):** All four conformance-test issues (#3124, #3273, #3385, #2841) now carry explicit references to commit `4cadec7f8` (PR #3991, `TestCodingAgentBackendConformance`). B2 behavioral conformance gaps (R-B/B2: env injection presence, resume-id inclusion, flag semantics) are noted where applicable. #3273 and #3385 identify specific B2 gaps; #2841's INDEPENDENT classification is unaffected by B2 tracking.
 
 ## Summary Statistics