claw/docs/superpowers/reports/2026-04-19-structured-fail-closed-improvement-roadmap-closure-report.md

Structured Fail-Closed Improvement Roadmap Closure Report

Date: 2026-04-19 Plan: docs/superpowers/plans/2026-04-19-structured-fail-closed-improvement-roadmap-plan.md Status: Closed

Scope

This roadmap handled the 48 structured fail-closed records from the reconciled 102 scene sweep.

The roadmap boundary explicitly excluded:

1. new scene families
2. G4/G5
3. login recovery
4. full host runtime transport
5. local document attachment runtime
6. scene_execution_board_2026-04-18.json updates
7. direct scene promotion
8. timeout implementation work
9. readiness gate relaxation

Those boundaries were preserved.

Completed Phases

Phase 0: Freeze Structured Fail-Closed Baseline

Completed.

Input baseline:

Status	Count
auto-pass	48
fail-closed-known	48
adjudicated-valid-host-bridge	4
source-unreadable	2
Total	102

Only fail-closed-known records entered the inventory phase.

Phase 1: Inventory and Gap Taxonomy

Completed.

Deliverables:

1. tests/fixtures/generated_scene/structured_fail_closed_inventory_2026-04-19.json
2. docs/superpowers/reports/2026-04-19-structured-fail-closed-inventory-report.md

Inventory summary:

Primary label	Count
enrichment_request_missing	23
export_plan_missing	12
local_doc_contract_missing	5
mode_request_contract_missing	4
single_request_enrichment_contract_missing	2
bootstrap_target_unresolved	2

The paginated_enrichment bucket was split into:

Primary label	Count
enrichment_request_missing	23
export_plan_missing	12

Phase 2: G3 Paginated Enrichment Recovery Slice

Completed as a bounded evidence-preservation slice.

The top repeated G3 gap was enrichment_request_missing, but there was not enough traceable evidence to safely synthesize missing enrichment requests without relaxing the G3 contract.

Instead, the implementation improved structured fail-closed observability by embedding contractSnapshot in pre-package fail-closed reports. This preserves the full SceneIr for future recovery analysis while keeping the result fail-closed.

Changed implementation:

1. src/generated_scene/generator.rs
2. tests/scene_generator_test.rs

This slice did not:

1. hardcode scene names
2. relax gates
3. convert incomplete G3 records into pass
4. change canonical G3 behavior

Phase 3: Small-Bucket Recovery Slice

Completed as inspection and explicit deferral.

Deliverable:

docs/superpowers/reports/2026-04-19-structured-fail-closed-small-bucket-decision-report.md

No small-bucket implementation slice was started because the largest remaining value stayed in the G3 structured fail-closed bucket, and the plan allowed at most one non-G3 bounded slice.

Phase 4: Bootstrap Target Isolation

Completed.

Deliverable:

docs/superpowers/reports/2026-04-19-structured-fail-closed-bootstrap-isolation-report.md

Bootstrap target records were kept out of G3 and small-bucket recovery. No login recovery or bootstrap auto-recovery was implemented.

Phase 5: Follow-Up Sweep and Coverage Delta

Completed.

Deliverables:

1. tests/fixtures/generated_scene/structured_fail_closed_improvement_followup_2026-04-19.json
2. docs/superpowers/reports/2026-04-19-structured-fail-closed-improvement-coverage-delta-report.md

Follow-up summary:

Status	Baseline	Follow-up	Delta
auto-pass	48	48	0
fail-closed-known	48	47	-1
adjudicated-valid-host-bridge	4	4	0
source-unreadable	2	3	+1

The follow-up did not improve pass count and exposed one additional timeout:

sweep-040-scene / 嘉峪关日报

Validation

The implementation slice was validated with:

1. cargo test --test scene_generator_test generator_blocks_incomplete_paginated_enrichment_workflow -- --nocapture
2. cargo test --test scene_generator_canonical_test -- --nocapture

Both passed.

The fixed 102 scene follow-up sweep completed and wrote:

tests/fixtures/generated_scene/structured_fail_closed_improvement_followup_2026-04-19.json

Closure Decision

This roadmap is closed.

It completed its required inventory, bounded implementation, small-bucket inspection, bootstrap isolation, follow-up sweep, and delta report.

The outcome is not a coverage increase. The outcome is better structured fail-closed evidence plus a measured follow-up baseline:

Metric	Result
auto-pass coverage	unchanged at 48/102
broad actionable count	100 -> 99
remaining structured fail-closed	47
remaining timeout	3

Next Inputs

The next work is outside this roadmap.

Highest-value inputs are:

1. bounded timeout regression diagnostic for sweep-040-scene
2. timeout diagnostic for the two persistent timeout records
3. a new bounded recovery roadmap for the remaining 34 paginated_enrichment structured fail-closed records

No next roadmap is started by this closure report.