fix: preserve requirement text and standalone document scope in v0.5.6

README.md

@@ -4,7 +4,7 @@
 
 **PRD·저장 가능한 작업 계획·요구사항 검증·Git 작업을 위한 선택형 스킬 모음.**
 
-![Version](https://img.shields.io/badge/version-0.5.5-6C5CE7?style=for-the-badge)
+![Version](https://img.shields.io/badge/version-0.5.6-6C5CE7?style=for-the-badge)
 ![Skills](https://img.shields.io/badge/core_skills-9-00B894?style=for-the-badge)
 ![Platform](https://img.shields.io/badge/platform-Codex-111827?style=for-the-badge)
 ![License](https://img.shields.io/badge/license-Apache--2.0-0984E3?style=for-the-badge)
@@ -120,6 +120,19 @@ release-readiness ──── 안전한 커밋·푸시·PR
 
 ---
 
+## v0.5.6: 요구사항 원문 보존과 단일 문서 범위 수정
+
+- 요구사항 importer가 `unit_price`, `ROUND_HALF_UP`, 곱셈 기호와 본문 내부 공백을 보존합니다. 서로 다른 본문을 같게 정규화해 이전 검증을 재사용하던 문제를 수정했습니다.
+- 단일 화면 명세·개발 인계는 요청한 파일과 형식에 맞춰 작성합니다. WIGTN 묶음 검증기는 해당 묶음과 의존 산출물을 요청한 경우에만 사용합니다.
+- 기존 Core 9개 스킬과 명시 호출형 `verified-delivery`는 유지합니다. Knowledge Wiki는 공동 버전 정책에 따른 manifest 변경만 포함합니다.
+- 이전에 가져온 원문이 손상됐다면 소스를 다시 import하고 검증해야 합니다. 서식 보존으로 자동 생성 ID나 이전 검증의 재사용 여부가 달라질 수 있습니다.
+
+[변경 내용과 검증 범위](docs/RELEASE-v0.5.6-KO.md) · [English release notes](docs/RELEASE-v0.5.6.md)
+
+이번 패치는 정확성과 요청 범위를 수정합니다. 수정본의 GPT-6 성능 향상을 주장하지 않습니다.
+
+---
+
 ## v0.5.5: WorkGraph 선택 조회와 실측 보고서
 
 - `inspect --task TASK-ID`로 지정한 작업·선행 작업과 연결된 기록을 조회합니다. 전체 검증·소스 변경 감지·릴리스 게이트와 생략 개수는 유지합니다.

docs/RELEASE-v0.5.6-KO.md

@@ -0,0 +1,39 @@
+# v0.5.6 — 요구사항 원문 보존과 단일 문서 범위 수정
+
+요구사항을 가져오는 과정에서 본문이 손상되던 문제와, 단일 화면 명세·개발 인계에
+전체 묶음 검증을 적용하던 문제를 수정했습니다.
+
+## 요구사항 원문 보존
+
+기존 importer는 본문의 밑줄·별표·backtick을 제거하고 공백을 합쳤습니다.
+예를 들어 `unit_price * quantity`가 `unitprice quantity`로 바뀔 수 있었습니다.
+이 정규화를 이전 근거의 재사용 판단에도 적용해 서로 다른 요구사항을 같은 것으로
+취급할 수 있었습니다.
+
+이제 본문의 식별자·연산자·기호·내부 공백을 보존합니다. 목록·표 입력, 음수,
+공백을 포함한 리터럴, 본문 변경 시 검증 무효화와 동일 본문의 검증 재사용에 대한
+회귀 검사를 추가했습니다. 기존 형식별 import와 재개 검사는 유지합니다.
+
+이미 저장된 원문이 손상됐다면 소스를 다시 import하고 검증해야 합니다.
+저장된 WorkGraph를 자동으로 다시 쓰지는 않습니다. 원문 보존으로 명시 ID가 없는
+형식의 자동 생성 ID가 바뀌거나, 서식 변경만으로도 이전 근거가 무효화될 수 있습니다.
+
+## 요청한 문서만 작성·검토
+
+screen-spec은 사용자가 요청한 산출물·경로·형식을 따릅니다. 기존 문서는 입력으로
+참조하고 템플릿은 필요할 때 사용합니다. 단일 채팅 답변이나 사용자 지정 Markdown은
+관련 요구사항을 직접 검토하며, 묶음 검증기를 통과시키려고 다른 문서를 추가하지 않습니다.
+
+번호가 붙은 WIGTN 산출물 묶음과 의존 문서가 요청 범위에 있는 경우에는 기존
+묶음 검증기를 사용합니다. 이 검증기의 구조 검사와 회귀 테스트는 유지합니다.
+
+## 호환성과 검증 범위
+
+- 기존 Core 9개 스킬과 명시 호출형 `verified-delivery`를 유지합니다.
+  실험 브랜치의 스킬 삭제는 포함하지 않습니다.
+- Knowledge Wiki는 공동 버전 정책에 따라 manifest 버전만 올립니다.
+- `bash scripts/validate.sh`의 저장소·플러그인·스킬·정적 회귀 검사가 통과했습니다.
+  기존 importer에서 버그를 재현했고 실제 WorkGraph CLI의 원문 보존도 확인했습니다.
+  릴리스 검사 로그는 [release-v0.5.6-validation.log](release-v0.5.6-validation.log)에 보존합니다.
+- 수정본의 추가 모델 벤치마크는 수행하지 않았습니다. GPT-6 속도·토큰·코딩 품질
+  향상을 주장하는 릴리스가 아닙니다.

docs/RELEASE-v0.5.6.md

@@ -0,0 +1,45 @@
+# v0.5.6 — Preserve requirement text and respect standalone document scope
+
+This patch fixes requirement-text corruption during import and unnecessary
+bundle validation for standalone screen specifications and developer handoffs.
+
+## Requirement imports
+
+The importer previously removed underscores, asterisks and backticks and
+collapsed whitespace in requirement bodies. For example, `unit_price * quantity`
+could become `unitprice quantity`, changing the meaning of the requirement.
+The same normalization could treat different requirement bodies as identical
+when deciding whether to retain earlier verification evidence.
+
+Requirement bodies now preserve these characters and internal whitespace.
+Regression checks cover list/table inputs, identifiers, operators, negative
+values, literal spaces, unchanged evidence reuse and invalidation after a
+meaningful text change. Existing format and resume checks remain in place.
+
+Re-import affected source documents and rerun verification to repair previously
+saved records. This update does not rewrite saved WorkGraphs automatically.
+Preserving text can change generated IDs in formats without explicit IDs and
+can conservatively invalidate evidence after formatting-only changes.
+
+## Standalone screen specifications and handoffs
+
+The screen-spec skill now follows the requested artifact, path and format.
+Existing documents are inputs, and templates are optional references.
+For standalone chat or custom Markdown output, the skill reviews the relevant
+requirements directly instead of running the numbered WIGTN bundle validator.
+That validator remains available when the numbered artifact set and its
+companion files are in scope; its existing structural checks are unchanged.
+
+## Compatibility and validation
+
+- All nine Core skills remain available, including explicit-only
+  `verified-delivery`. Experimental skill removal is not part of this release.
+- Knowledge Wiki receives only the matching manifest version under the
+  repository's shared-version policy.
+- `bash scripts/validate.sh` passed repository, plugin, skill and deterministic
+  regression validation. A negative control reproduced the old importer defect,
+  and a WorkGraph CLI smoke check preserved literal requirement text. The
+  release log is stored as `release-v0.5.6-validation.log`.
+- These are correctness and scope fixes. The patched release has not undergone
+  a new model benchmark, and no GPT-6 speed, token or coding-quality improvement
+  is claimed.

docs/release-v0.5.6-validation.log

@@ -0,0 +1,141 @@
+Repository contract: PASS
+Plugin validation passed: /private/tmp/wigtn-release-v0.5.6-20260920/plugins/wigtn-plugins-with-codex
+Plugin validation passed: /private/tmp/wigtn-release-v0.5.6-20260920/plugins/wigtn-knowledge-wiki
+Skill is valid!
+Skill is valid!
+Skill is valid!
+Skill is valid!
+Skill is valid!
+Skill is valid!
+Skill is valid!
+Skill is valid!
+Skill is valid!
+Skill is valid!
+Verified-delivery routing: PASS (10 contracts)
+External paired results: PASS (8 confirmatory runs, 4 pairs, 2 exploratory replication runs)
+FeatureBench selection: PASS (4 pilot, 8 frozen candidates)
+featurebench-results: ok (4 pairs, 0 eligible positive, 0 replicated positive)
+Knowledge wiki checks: PASS
+run-evals.sh is the compatibility entrypoint for deterministic static contracts.
+Use scripts/run-behavior-evals.sh --execute for model behavior smoke runs.
+Lexical trigger fixture: PASS (49 cases; model routing not tested)
+Plugin resource integrity: PASS (40 markdown files)
+Visual contracts: PASS (diagram/html/presentation-routing)
+Evidence contract: PASS (10 cases)
+Evidence inspector: PASS (health/drift)
+Project context: PASS (valid/invalid)
+PRD contracts: PASS (4 cases)
+Screen contract: PASS (selected/closure/multilingual-ia/bundle/anchor/requirement)
+Release state: PASS (paths, bounded diffs, whitespace checks)
+Requirement import: PASS (3 formats + literal preservation + resume/drift)
+test_acceptance_section_preserves_underscore_ids (__main__.SymbolicIDs.test_acceptance_section_preserves_underscore_ids) ... ok
+test_contracts_accept_same_symbolic_id (__main__.SymbolicIDs.test_contracts_accept_same_symbolic_id) ... ok
+test_explicit_import_preserves_complete_ids (__main__.SymbolicIDs.test_explicit_import_preserves_complete_ids) ... ok
+test_markdown_lists_and_checkboxes_remain_supported (__main__.SymbolicIDs.test_markdown_lists_and_checkboxes_remain_supported) ... ok
+test_prose_and_incomplete_ids_are_not_requirements (__main__.SymbolicIDs.test_prose_and_incomplete_ids_are_not_requirements) ... ok
+test_workgraph_import_plan_and_source_drift (__main__.SymbolicIDs.test_workgraph_import_plan_and_source_drift) ... ok
+
+----------------------------------------------------------------------
+Ran 6 tests in 0.454s
+
+OK
+WorkGraph contract: PASS (67 deterministic cases)
+Evaluation schedule: PASS (deterministic/block-balanced/paired)
+Ablation builder: PASS (core4/placebo4/full8/full9)
+Ordinary corpus: PASS (12 tasks, 3 languages, fault/reference)
+Ordinary scorer: PASS (positive + hidden-regression rejection)
+Explicit pricing: PASS (unpriced/partial/explicit/nonfinite)
+Codex usage: PASS (JSONL fields + API-equivalent cost)
+test_absent_optional_is_unknown (__main__.UsageMissing.test_absent_optional_is_unknown) ... ok
+test_cli_has_no_assumed_prices (__main__.UsageMissing.test_cli_has_no_assumed_prices) ... ok
+test_explicit_zero_is_known (__main__.UsageMissing.test_explicit_zero_is_known) ... ok
+test_impossible_breakdown_rejected (__main__.UsageMissing.test_impossible_breakdown_rejected) ... ok
+test_invalid_optional_counters_rejected (__main__.UsageMissing.test_invalid_optional_counters_rejected) ... ok
+test_invalid_required_counters_rejected (__main__.UsageMissing.test_invalid_required_counters_rejected) ... ok
+test_missing_cache_prevents_cache_adjusted_price (__main__.UsageMissing.test_missing_cache_prevents_cache_adjusted_price) ... ok
+test_missing_one_turn_propagates (__main__.UsageMissing.test_missing_one_turn_propagates) ... ok
+test_summary_preserves_unknown_without_crashing (__main__.UsageMissing.test_summary_preserves_unknown_without_crashing) ... ok
+
+----------------------------------------------------------------------
+Ran 9 tests in 0.080s
+
+OK
+Token ablation scorer: PASS (positive + canonical regression)
+Eval packet: PASS (sanitize/hash/tamper)
+Research harness: PASS (8 acceptance, 4 implement, 10 release, 12 WorkGraph + resume, 12 ordinary)
+test_activation_needs_real_trace_reference (__main__.EvidenceTests.test_activation_needs_real_trace_reference) ... ok
+test_adjudicated_packet_passes (__main__.EvidenceTests.test_adjudicated_packet_passes) ... ok
+test_changed_events_rejected (__main__.EvidenceTests.test_changed_events_rejected) ... ok
+test_changed_implementation_is_rejected (__main__.EvidenceTests.test_changed_implementation_is_rejected) ... ok
+test_early_stop_retains_observed_consumption (__main__.EvidenceTests.test_early_stop_retains_observed_consumption) ... ok
+test_empty_schedule_rejected (__main__.EvidenceTests.test_empty_schedule_rejected) ... ok
+test_extra_work_file_rejected (__main__.EvidenceTests.test_extra_work_file_rejected) ... ok
+test_failed_turn_consumption_is_retained (__main__.EvidenceTests.test_failed_turn_consumption_is_retained) ... ok
+test_failure_cost_retained (__main__.EvidenceTests.test_failure_cost_retained) ... ok
+test_input_snapshot_mutation_rejected (__main__.EvidenceTests.test_input_snapshot_mutation_rejected) ... ok
+test_invalid_turn_telemetry_is_not_complete (__main__.EvidenceTests.test_invalid_turn_telemetry_is_not_complete) ... ok
+test_missing_seal_rejected (__main__.EvidenceTests.test_missing_seal_rejected) ... ok
+test_missing_usage_leaves_incomplete_total (__main__.EvidenceTests.test_missing_usage_leaves_incomplete_total) ... ok
+test_new_contract_cases_and_complete_criteria (__main__.EvidenceTests.test_new_contract_cases_and_complete_criteria) ... ok
+test_offline_runner_to_scorer_integration (__main__.EvidenceTests.test_offline_runner_to_scorer_integration) ... ok
+test_oracle_catches_vacuous_test_and_passes_fix (__main__.EvidenceTests.test_oracle_catches_vacuous_test_and_passes_fix) ... ok
+test_review_pending_is_not_pass (__main__.EvidenceTests.test_review_pending_is_not_pass) ... ok
+test_skill_name_without_body_is_not_activation (__main__.EvidenceTests.test_skill_name_without_body_is_not_activation) ... ok
+test_stale_review_rejected (__main__.EvidenceTests.test_stale_review_rejected) ... ok
+test_symlinked_evidence_rejected (__main__.EvidenceTests.test_symlinked_evidence_rejected) ... ok
+test_unknown_reasoning_is_null (__main__.EvidenceTests.test_unknown_reasoning_is_null) ... ok
+
+----------------------------------------------------------------------
+Ran 21 tests in 1.845s
+
+OK
+test_corrupt_evidence_withholds_eligible_tasks (__main__.Inspect.test_corrupt_evidence_withholds_eligible_tasks) ... ok
+test_drift_propagates_to_dependent_tasks_without_mutation (__main__.Inspect.test_drift_propagates_to_dependent_tasks_without_mutation) ... ok
+test_fresh_readonly_and_no_saved_command_execution (__main__.Inspect.test_fresh_readonly_and_no_saved_command_execution) ... ok
+test_invalid_global_evidence_still_blocks_selection (__main__.Inspect.test_invalid_global_evidence_still_blocks_selection) ... ok
+test_invalid_graph_fails_closed (__main__.Inspect.test_invalid_graph_fails_closed) ... ok
+test_invalid_project_withholds_eligible_tasks (__main__.Inspect.test_invalid_project_withholds_eligible_tasks) ... ok
+test_missing_source_has_explicit_drift (__main__.Inspect.test_missing_source_has_explicit_drift) ... ok
+test_repeated_selection_deduplicates (__main__.Inspect.test_repeated_selection_deduplicates) ... ok
+test_selected_drift_is_visible_and_saved_revision_is_not_preview (__main__.Inspect.test_selected_drift_is_visible_and_saved_revision_is_not_preview) ... ok
+test_selection_omits_dependents_but_keeps_global_summary_and_gates (__main__.Inspect.test_selection_omits_dependents_but_keeps_global_summary_and_gates) ... ok
+test_selection_preserves_dependency_records_and_linked_context (__main__.Inspect.test_selection_preserves_dependency_records_and_linked_context) ... ok
+test_transitive_dependencies_and_shared_checks (__main__.Inspect.test_transitive_dependencies_and_shared_checks) ... ok
+test_unknown_selection_fails (__main__.Inspect.test_unknown_selection_fails) ... ok
+test_unrelated_source_drift_is_not_hidden (__main__.Inspect.test_unrelated_source_drift_is_not_hidden) ... ok
+
+----------------------------------------------------------------------
+Ran 14 tests in 2.059s
+
+OK
+test_absolute_inside_allowed_outside_rejected_without_mutation (__main__.ImportRoot.test_absolute_inside_allowed_outside_rejected_without_mutation) ... ok
+test_relative_source_uses_root_and_preserves_dry_run (__main__.ImportRoot.test_relative_source_uses_root_and_preserves_dry_run) ... ok
+test_symlink_escape_rejected_without_mutation (__main__.ImportRoot.test_symlink_escape_rejected_without_mutation) ... ok
+
+----------------------------------------------------------------------
+Ran 3 tests in 0.517s
+
+OK
+test_dry_run_does_not_create_lock_or_change_state (__main__.Concurrency.test_dry_run_does_not_create_lock_or_change_state) ... ok
+test_killed_writer_releases_lock (__main__.Concurrency.test_killed_writer_releases_lock) ... ok
+test_no_revision_second_writer_merges_latest (__main__.Concurrency.test_no_revision_second_writer_merges_latest) ... ok
+test_same_revision_second_writer_rejected (__main__.Concurrency.test_same_revision_second_writer_rejected) ... ok
+
+----------------------------------------------------------------------
+Ran 4 tests in 2.066s
+
+OK
+test_invalid_context_preserves_graph_for_preview_and_apply (__main__.ProjectGuard.test_invalid_context_preserves_graph_for_preview_and_apply) ... ok
+
+----------------------------------------------------------------------
+Ran 1 test in 1.752s
+
+OK
+Static contracts: PASS
+Static validation: PASS
+
+Additional release checks:
+PASS: new literal regression rejected the released importer and passed with the patch.
+CLI smoke setup initially omitted the required init; corrected to init -> import without product changes.
+PASS: WorkGraph CLI init -> import preserves identifiers and operator end to end
+PASS: only intended patch paths changed; nine skills retained; verified-delivery unchanged; manifest changes version-only

plugins/wigtn-knowledge-wiki/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "wigtn-knowledge-wiki",
-  "version": "0.5.5",
+  "version": "0.5.6",
   "description": "Opt-in, gated Codex session knowledge capture for an explicitly authorized personal wiki namespace.",
   "author": {
     "name": "WIGTN"

plugins/wigtn-plugins-with-codex/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "wigtn-plugins-with-codex",
-  "version": "0.5.5",
+  "version": "0.5.6",
   "description": "Selective product contracts for Codex: compact PRDs, saved WorkGraphs, scoped screen artifacts, acceptance evidence, explicit verified delivery, and safe release actions.",
   "author": {
     "name": "WIGTN"

plugins/wigtn-plugins-with-codex/scripts/import-requirements.py

@@ -44,8 +44,10 @@ def detect_format(path: Path, text: str) -> str:
 
 
 def clean_text(value: str) -> str:
-    value = re.sub(r"[*_`]+", "", value)
-    return re.sub(r"\s+", " ", value).strip(" |:-")
+    # Markdown markers can also be identifier characters, operators or literal
+    # data. Preserve requirement content; structural prefixes are parsed by the
+    # caller. Formatting changes may conservatively invalidate old evidence.
+    return value.strip()
 
 
 def derived_id(prefix: str, source: str, used: set[str]) -> str: