@dzhechkov/keysarium 1.7.14 → 1.7.18

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (84) hide show
  1. package/.dz-manifest.json +215 -55
  2. package/README.md +35 -6
  3. package/README.ru.md +36 -16
  4. package/package.json +3 -3
  5. package/sbom.json +456 -56
  6. package/src/cli.js +32 -2
  7. package/src/commands/doctor.js +25 -1
  8. package/src/commands/init.js +46 -12
  9. package/src/utils.js +1 -0
  10. package/templates/.agents/skills/casarium/SKILL.md +29 -0
  11. package/templates/.claude/agents/code-goal-planner.md +9 -0
  12. package/templates/.claude/commands/analyst-manual.md +4 -2
  13. package/templates/.claude/commands/casarium.md +3 -1
  14. package/templates/.claude/commands/cjm-prototype.md +2 -2
  15. package/templates/.claude/commands/research.md +4 -1
  16. package/templates/.claude/rules/checkpoint-protocol.md +2 -1
  17. package/templates/.claude/rules/research-quality.md +11 -6
  18. package/templates/.claude/shards/phase-2-research.shard.md +12 -8
  19. package/templates/.claude/shards/phase-25-cjm.shard.md +5 -3
  20. package/templates/.claude/skills/analyst-manual-full/SKILL.md +65 -37
  21. package/templates/.claude/skills/feature-adr/SKILL.md +17 -15
  22. package/templates/.claude/skills/feature-adr/modules/03-adr.md +1 -1
  23. package/templates/.claude/skills/feature-adr/modules/03.5-ideation-swarm.md +1 -1
  24. package/templates/.claude/skills/feature-adr/modules/06-implementation-plan.md +2 -2
  25. package/templates/.claude/skills/feature-adr/modules/08-qe.md +1 -1
  26. package/templates/.claude/skills/feature-adr/modules/09-fleet-qe.md +12 -12
  27. package/templates/.claude/skills/feature-adr/references/agentic-qe/LICENSE.agentic-qe +21 -0
  28. package/templates/.claude/skills/feature-adr/references/agentic-qe/README.md +20 -16
  29. package/templates/.claude/skills/feature-adr/references/agentic-qe/UPSTREAM-MANIFEST.json +246 -0
  30. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-brutal-honesty-review/README.md +218 -0
  31. package/templates/.claude/skills/feature-adr/references/agentic-qe/{brutal-honesty-review.md → qe-ref-brutal-honesty-review/REFERENCE.md} +14 -1
  32. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-brutal-honesty-review/resources/assessment-rubrics.md +295 -0
  33. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-brutal-honesty-review/resources/review-template.md +102 -0
  34. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-brutal-honesty-review/schemas/output.json +291 -0
  35. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-brutal-honesty-review/scripts/assess-code.sh +179 -0
  36. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-brutal-honesty-review/scripts/assess-tests.sh +223 -0
  37. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-brutal-honesty-review/scripts/validate-config.json +34 -0
  38. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-enterprise-integration-testing/REFERENCE.md +758 -0
  39. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-enterprise-integration-testing/evals/enterprise-integration-testing.yaml +158 -0
  40. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-enterprise-integration-testing/schemas/output.json +74 -0
  41. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-enterprise-integration-testing/scripts/validate-config.json +25 -0
  42. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qcsd-ideation-swarm/REFERENCE.md +226 -0
  43. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qcsd-ideation-swarm/evals/qcsd-ideation-swarm.yaml +138 -0
  44. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qcsd-ideation-swarm/schemas/output.json +568 -0
  45. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qcsd-ideation-swarm/scripts/validate-config.json +25 -0
  46. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qcsd-ideation-swarm/steps/01-flag-detection.md +57 -0
  47. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qcsd-ideation-swarm/steps/02-core-agents.md +29 -0
  48. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qcsd-ideation-swarm/steps/03-batch1-results.md +15 -0
  49. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qcsd-ideation-swarm/steps/04-conditional-agents.md +23 -0
  50. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qcsd-ideation-swarm/steps/05-decision-synthesis.md +29 -0
  51. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qcsd-ideation-swarm/steps/06-report-generation.md +18 -0
  52. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qcsd-ideation-swarm/steps/07-learning-persistence.md +29 -0
  53. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qcsd-ideation-swarm/steps/08-final-output.md +18 -0
  54. package/templates/.claude/skills/feature-adr/references/agentic-qe/{qe-coverage-analysis.md → qe-ref-qe-coverage-analysis/REFERENCE.md} +42 -3
  55. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qe-coverage-analysis/evals/qe-coverage-analysis.yaml +494 -0
  56. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qe-coverage-analysis/schemas/output.json +286 -0
  57. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qe-coverage-analysis/scripts/validate-config.json +42 -0
  58. package/templates/.claude/skills/feature-adr/references/agentic-qe/{qe-requirements-validation.md → qe-ref-qe-requirements-validation/REFERENCE.md} +13 -3
  59. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qe-requirements-validation/evals/qe-requirements-validation.yaml +598 -0
  60. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qe-requirements-validation/schemas/output.json +587 -0
  61. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-qe-requirements-validation/scripts/validate-config.json +37 -0
  62. package/templates/.claude/skills/feature-adr/references/agentic-qe/{regression-testing.md → qe-ref-regression-testing/REFERENCE.md} +30 -49
  63. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-regression-testing/evals/regression-testing.yaml +504 -0
  64. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-regression-testing/schemas/output.json +311 -0
  65. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-regression-testing/scripts/validate-config.json +50 -0
  66. package/templates/.claude/skills/feature-adr/references/agentic-qe/{risk-based-testing.md → qe-ref-risk-based-testing/REFERENCE.md} +23 -52
  67. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-risk-based-testing/evals/risk-based-testing.yaml +141 -0
  68. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-risk-based-testing/schemas/output.json +480 -0
  69. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-risk-based-testing/scripts/validate-config.json +25 -0
  70. package/templates/.claude/skills/feature-adr/references/agentic-qe/{shift-left-testing.md → qe-ref-shift-left-testing/REFERENCE.md} +33 -88
  71. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-shift-left-testing/evals/shift-left-testing.yaml +145 -0
  72. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-shift-left-testing/schemas/output.json +459 -0
  73. package/templates/.claude/skills/feature-adr/references/agentic-qe/qe-ref-shift-left-testing/scripts/validate-config.json +25 -0
  74. package/templates/CLAUDE.md +3 -1
  75. package/templates/docs/feature-adr/04-admin-guide.md +1 -1
  76. package/templates/docs/improvement-directions.md +3 -3
  77. package/templates/docs/keysarium/02-admin-guide.md +2 -2
  78. package/templates/docs/keysarium/03-user-guide.md +3 -3
  79. package/templates/docs/keysarium/05-architecture.md +9 -8
  80. package/templates/docs/keysarium/06-user-flows.md +4 -4
  81. package/templates/lib/platform-templates/copilot.md +1 -1
  82. package/templates/lib/platform-templates/opencode.md +1 -1
  83. package/templates/.claude/skills/feature-adr/references/agentic-qe/enterprise-integration-testing.md +0 -111
  84. package/templates/.claude/skills/feature-adr/references/agentic-qe/qcsd-ideation-swarm.md +0 -126
@@ -35,7 +35,7 @@ Scripts и references для Phase 2 находятся в:
35
35
 
36
36
  **Контрольные точки:** После каждой фазы скилл останавливается для подтверждения. Пользователь может скорректировать направление или продолжить.
37
37
 
38
- **Cryptographic Verification:** Ed25519 подписи для верификации источников (через goap-research-ed25519).
38
+ **Проверка происхождения:** Research фиксирует, какие источники были прочитаны и какие утверждения ими поддержаны. Ed25519 проверяет подпись издателя только при наличии настоящих ключей издателя и источника с подписанной выдачей. Подпись подтверждает происхождение конкретных байтов, а не истинность утверждения.
39
39
 
40
40
  ## Workflow Architecture
41
41
 
@@ -52,9 +52,9 @@ Scripts и references для Phase 2 находятся в:
52
52
  │ → Output: Task Brief │
53
53
  │ → ⏸️ CHECKPOINT 1: "Подтвердите Task Brief" │
54
54
  │ ↓ │
55
- │ Phase 2: RESEARCH (Ed25519 Verified) │
55
+ │ Phase 2: RESEARCH (GOAP + статус доказательств) │
56
56
  │ → Загрузи: ../goap-research-ed25519/SKILL.md │
57
- │ → Output: Verified Research Findings │
57
+ │ → Output: Research Findings + Verification Status │
58
58
  │ → ⏸️ CHECKPOINT 2: "Подтвердите findings" │
59
59
  │ ↓ │
60
60
  │ Phase 3: SOLVE │
@@ -105,57 +105,82 @@ Scripts и references для Phase 2 находятся в:
105
105
  [показать Task Brief]
106
106
 
107
107
  Варианты действий:
108
- • "ок" — перейти к Verified Research (moderate mode)
109
- • "режим strict" — research с порогом 0.95
110
- • "режим paranoid" — research с порогом 0.99
108
+ • "ок" — перейти к Research (moderate mode)
111
109
  • "скорректируй X" — внести изменения
112
110
 
113
111
  Что выберете?
114
112
  ═══════════════════════════════════════════════════════════════
115
113
  ```
116
114
 
115
+ Показывай дополнительные варианты `режим strict` и `режим paranoid` только после проверки **обоих** фактов для выбранного класса источников: `issuer_keys_available` (реальные закреплённые ключи издателей) и `source_class_verified` (источники действительно отдают подписанные данные). Если пользователь просит один из этих режимов без подтверждения хотя бы одного факта, назови недостающий факт и остановись на CHECKPOINT 1. Предложи `moderate` отдельно; не подменяй выбор молча. Для обычных PubMed/PMC/DOI/WHO эти факты не установлены.
116
+
117
+ | Ключи издателя | Подписанная выдача | Можно предложить `strict`/`paranoid`? |
118
+ |---|---|---|
119
+ | Нет | Нет | Нет; штатный выбор — `moderate`. |
120
+ | Да | Нет | Нет; назови недостающую подписанную выдачу. |
121
+ | Нет | Да | Нет; назови недостающие ключи издателя. |
122
+ | Да | Да | Да, после явного выбора пользователя. |
123
+
117
124
  ### Phase 2: Research (goap-research-ed25519)
118
125
 
119
126
  **Загрузи:** Прочитай `../goap-research-ed25519/SKILL.md` и выполни по его инструкциям.
120
127
 
121
128
  **Verification Modes (выбор на CHECKPOINT 1):**
122
129
 
123
- | Mode | Threshold | Команда на checkpoint |
124
- |------|-----------|----------------------|
125
- | `moderate` | 0.85 | "ок" или "продолжай" |
126
- | `strict` | 0.95 | "режим strict" |
127
- | `paranoid` | 0.99 | "режим paranoid" |
130
+ | Mode | Threshold | Условие |
131
+ |------|-----------|---------|
132
+ | `moderate` | 0.85 | Штатный режим по "ок" или "продолжай"; неподписанные факты остаются явно отмеченными. |
133
+ | `strict` | 0.95 | Только когда установлены оба capability-факта и пользователь выбрал режим. |
134
+ | `paranoid` | 0.99 | То же условие; более сильные требования к избыточности источников. |
135
+
136
+ Перед исследованием передай установленные capability-факты в `GOAPResearchPlanner`; не выставляй флаги `True` по имени сайта, наличию URL или желанию получить высокий confidence. Обрабатывай ответ как `ResearchPlan` **или** `PlanNotFound`, не как `None`:
137
+
138
+ - `PlanNotFound` с `GOAL_UNREACHABLE`: покажи verdict, причину и недостающие факты; остановись на research checkpoint. Увеличение бюджета поиска не исправит недоступные подписи.
139
+ - `PlanNotFound` с `SEARCH_EXHAUSTED`: покажи verdict и причину; остановись на research checkpoint. Можно предложить ограниченный повторный поиск, но не объявлять план найденным.
140
+ - Только `ResearchPlan` разрешает перейти к сбору находок; Phase 3: Solve доступна после успешного CHECKPOINT 2. `ResearchPlan` сам по себе не доказывает подписи или истинность фактов.
128
141
 
129
- **Output — Verified Research Findings:**
142
+ Если получен `PlanNotFound`, вместо шаблона «Research Findings Ready» покажи отдельный CHECKPOINT 2: «Research остановлен», фактический verdict, сообщение planner и недостающие capability-факты либо исчерпанный бюджет поиска. Предложи новый явный выбор: перейти на `moderate` при `GOAL_UNREACHABLE` из-за недоступной подписи, исправить интеграцию подписанной выдачи или сузить цель/повторить поиск при `SEARCH_EXHAUSTED`. Не переходи к Solve и не заполняй успешный research-шаблон на этом пути.
143
+
144
+ Для каждого использованного факта запиши результат `Ed25519Verifier.verify_fact`: `verified`, `trust_class` и причину отказа. Отдельно запиши результат `verify_citation_chain`; целостность цепочки исследователя не превращает `SELF_ATTESTED` в подпись издателя. Считай `issuer-verified` только факты с `verified=true` и `trust_class=ISSUER_SIGNED`; `SELF_ATTESTED` и `UNVERIFIED` показывай отдельно. Confidence, наличие DOI/URL и подпись исследователя не повышают класс до `ISSUER_SIGNED`. Если проверка не запускалась, напиши «не проверено», не ноль успешных или зелёный статус.
145
+
146
+ **Output — Research Findings:**
130
147
  ```markdown
131
- ## Verified Research Findings
148
+ ## Research Findings
132
149
 
133
150
  ### Verification Status
134
151
  - **Mode:** [selected_mode] ([threshold] threshold)
135
- - **Chain Integrity:** ✅ VERIFIED (ed25519)
136
- - **Unsigned Claims:** 0
137
- - **Total Citations:** [число]
152
+ - **Source access:** [сколько источников прочитано; сколько только из выдачи; сколько не открыто]
153
+ - **Issuer provenance:** [issuer-verified X/N; SELF_ATTESTED Y/N; UNVERIFIED Z/N из результатов verify_fact]
154
+ - **Citation-chain integrity:** [verified / failed / not checked по verify_citation_chain]
155
+ - **Claims without issuer verification:** [идентификаторы и причины; если нет — 0 по результатам]
156
+ - **Total citations:** [из списка фактически использованных ссылок]
138
157
 
139
- ### Verified Findings
140
- [Findings с signed citations]
158
+ ### Findings
159
+ [Утверждение, ссылка, evidence class и его фактический trust class]
141
160
 
142
161
  ### Confidence Assessment
143
- - **Cryptographically Verified (≥0.95):** [список]
144
- - **Cross-Reference Verified (0.85-0.95):** [список]
162
+ - **Confidence:** [оценки с отдельным основанием; порог не заменяет подпись]
145
163
 
146
164
  ### Citation Chain
147
- [Ed25519 verified chain — см. Appendix]
165
+ [Что именно проверено, каким ключом и с каким результатом; если не запускалось — not checked]
148
166
  ```
149
167
 
168
+ | Состояние результатов | Допустимая формулировка |
169
+ |---|---|
170
+ | Нет `ISSUER_SIGNED` с `verified=true` | «Источники процитированы; криптографическая проверка издателя не установлена». Укажи `SELF_ATTESTED` и `UNVERIFIED` отдельно. |
171
+ | Часть фактов issuer-verified | «Подпись издателя подтверждена для X из N фактов». Перечисли остальные факты и причины; общего «все проверено» нет. |
172
+ | Все факты issuer-verified | «Подпись издателя подтверждена для N из N фактов» **только после** фактического `verify_fact` для каждого факта. Целостность цепочки заявляй лишь при отдельном успешном `verify_citation_chain`; истинность фактов из подписи не следует. |
173
+
150
174
  **⏸️ CHECKPOINT 2:**
151
175
  ```
152
176
  ═══════════════════════════════════════════════════════════════
153
- ⏸️ CHECKPOINT 2: Verified Research Complete
177
+ ⏸️ CHECKPOINT 2: Research Findings Ready
154
178
 
155
179
  **Verification Status:**
156
- - Mode: [strict/moderate/paranoid]
157
- - Unsigned Claims: 0 ✅
158
- - Citation Chain: VERIFIED (Ed25519) ✅
180
+ - Mode: [фактически выбранный режим]
181
+ - Issuer provenance: [X/N issuer-verified; Y self-attested; Z unverified]
182
+ - Citation-chain integrity: [verified / failed / not checked]
183
+ - Неустановленные элементы: [факты и причины]
159
184
 
160
185
  **Ключевые находки:**
161
186
  [краткое summary]
@@ -165,7 +190,7 @@ Scripts и references для Phase 2 находятся в:
165
190
 
166
191
  Варианты действий:
167
192
  • "ок" — перейти к Solution
168
- • "исследуй глубже X" — дополнительный verified research
193
+ • "исследуй глубже X" — дополнительный research с новыми источниками
169
194
  • "повысь confidence для Y" — найти больше источников
170
195
 
171
196
  Что выберете?
@@ -210,8 +235,9 @@ Scripts и references для Phase 2 находятся в:
210
235
  [краткое summary]
211
236
 
212
237
  **Data Quality:**
213
- - Все рекомендации основаны на verified research ✅
214
- - Citation chain integrity: VERIFIED (Ed25519) ✅
238
+ - Основание рекомендаций: [перечень подтверждённых и неподтверждённых находок]
239
+ - Issuer provenance: [X/N фактов, по результатам verify_fact]
240
+ - Citation-chain integrity: [отдельный результат verify_citation_chain]
215
241
 
216
242
  Варианты действий:
217
243
  • "ок" — создать все документы
@@ -228,10 +254,11 @@ Scripts и references для Phase 2 находятся в:
228
254
  ## Strategic Analysis: [Название задачи]
229
255
 
230
256
  ### Executive Summary
231
- ### Verification Status (Ed25519)
257
+ ### Verification Status
232
258
  - Research Mode: [mode]
233
- - All Claims Verified: ✅
234
- - Citation Chain: CRYPTOGRAPHICALLY VERIFIED ✅
259
+ - Issuer provenance: [X/N фактов; причины для остальных]
260
+ - Citation-chain integrity: [verified / failed / not checked]
261
+ - Source support: [какие утверждения подтверждены прочитанными источниками; какие нет]
235
262
 
236
263
  ### Key Research Insights
237
264
  ### Recommended Strategy
@@ -258,17 +285,18 @@ Scripts и references для Phase 2 находятся в:
258
285
  | Документ | Файлы |
259
286
  |----------|-------|
260
287
  | Task Brief | `01_task_brief.md`, `01_task_brief.docx` |
261
- | Verified Research | `02_research_findings.md`, `02_research_findings.docx` |
288
+ | Research Findings | `02_research_findings.md`, `02_research_findings.docx` |
262
289
  | Solution | `03_solution.md`, `03_solution.docx` |
263
290
  | Final Summary | `04_final_summary.md`, `04_final_summary.docx` |
264
291
 
265
292
  ## Quality Standards
266
293
 
267
294
  **Anti-Hallucination Checks:**
268
- - [ ] 100% claims имеют citations
269
- - [ ] Citation chain cryptographically verified (Ed25519)
270
- - [ ] Нет утверждений без DOI/URL
271
- - [ ] Confidence levels для всех findings
295
+ - [ ] Для каждого существенного утверждения указан источник либо явно сказано, что подтверждения нет.
296
+ - [ ] Статус `ISSUER_SIGNED` взят из результата `verify_fact`, а не из confidence, режима или подписи исследователя.
297
+ - [ ] Статус цепочки взят отдельно из `verify_citation_chain`; если проверка не запускалась, это написано.
298
+ - [ ] Утверждения без issuer verification перечислены, а `SELF_ATTESTED` не названы подписями издателя.
299
+ - [ ] Confidence и ограничения источника показаны отдельно от криптографического происхождения.
272
300
 
273
301
  **Completeness:**
274
302
  - [ ] Все 9 модулей problem-solver пройдены
@@ -278,7 +306,7 @@ Scripts и references для Phase 2 находятся в:
278
306
 
279
307
  **Required skills** (installed as sibling skill directories alongside this skill):
280
308
  - `explore` — task clarification and brief generation
281
- - `goap-research-ed25519` — verified research with Ed25519 (includes scripts/ and references/)
309
+ - `goap-research-ed25519` — source-aware research with optional issuer-signature verification (includes scripts/ and references/)
282
310
  - `problem-solver-enhanced` — 9-module problem solving with TRIZ
283
311
 
284
312
  **Runtime:**
@@ -60,17 +60,19 @@ Trigger on:
60
60
  │ ├── adr-template.md ← ADR document template
61
61
  │ ├── c4-template.md ← C4 diagram templates
62
62
  │ ├── qe-checklist.md ← Quality engineering checklist
63
- │ └── agentic-qe/ ← Integrated agentic-qe skills (15 total)
63
+ │ └── agentic-qe/ ← 15 QE references (8 pinned protocols)
64
64
  │ ├── README.md ← Skill mapping, modes, install guide
65
- │ ├── shift-left-testing.md ← Step 3: ADR testability (Tier 3)
66
- │ ├── qcsd-ideation-swarm.md ← Step 3.5: Quality swarm (Tier 3)
65
+ │ ├── UPSTREAM-MANIFEST.json ← Pinned upstream paths and hashes
66
+ │ ├── LICENSE.agentic-qe ← Upstream MIT notice
67
+ │ ├── qe-ref-shift-left-testing/REFERENCE.md ← Step 3: ADR testability (Tier 3)
68
+ │ ├── qe-ref-qcsd-ideation-swarm/REFERENCE.md ← Step 3.5: Quality swarm (Tier 3)
67
69
  │ ├── code-goal-planner.md ← Step 6: SPARC-GOAP agent
68
- │ ├── brutal-honesty-review.md ← Step 8: Code review (Tier 2)
69
- │ ├── qe-requirements-validation.md ← Step 9: Traceability (Tier 3)
70
- │ ├── risk-based-testing.md ← Step 9: Risk scoring (Tier 3)
71
- │ ├── enterprise-integration-testing.md ← Step 9: Integration (Tier 3, condensed)
72
- │ ├── regression-testing.md ← Step 9: Regression (Tier 3)
73
- │ ├── qe-coverage-analysis.md ← Step 9: Coverage (Tier 3)
70
+ │ ├── qe-ref-brutal-honesty-review/REFERENCE.md ← Step 8: Code review (Tier 2)
71
+ │ ├── qe-ref-qe-requirements-validation/REFERENCE.md ← Step 9: Traceability (Tier 3)
72
+ │ ├── qe-ref-risk-based-testing/REFERENCE.md ← Step 9: Risk scoring (Tier 3)
73
+ │ ├── qe-ref-enterprise-integration-testing/REFERENCE.md ← Step 9: Integration
74
+ │ ├── qe-ref-regression-testing/REFERENCE.md ← Step 9: Regression (Tier 3)
75
+ │ ├── qe-ref-qe-coverage-analysis/REFERENCE.md ← Step 9: Coverage (Tier 3)
74
76
  │ ├── chaos-engineering-resilience.md ← Extended: Chaos (Tier 3)
75
77
  │ ├── security-testing.md ← Extended: OWASP (Tier 3)
76
78
  │ ├── performance-testing.md ← Extended: Load/stress (Tier 3)
@@ -649,17 +651,17 @@ statusline/JSON and its ETA keep their existing behavior. Do not start a termina
649
651
 
650
652
  ### What changes with `--full-qe`
651
653
 
652
- Full agentic-qe protocols for the same 9 core skills. No new agents, just deeper methodology.
654
+ The eight vendored core skills already provide pinned full protocols in Reference Mode. Direct Mode adds the installed aqe runtime, its MCP pattern memory, and the steps explicitly listed below; it does not make the bundled references more complete.
653
655
 
654
656
  | Step | Reference Mode (default) | Direct Mode (`--full-qe`) |
655
657
  |------|--------------------------|---------------------------|
656
658
  | Step 0 | Standard routing | + pattern recall → `{LEARNED_PATTERNS}` |
657
- | Step 3 | Condensed shift-left protocol | Full Level 1-4 shift-left with BDD generators |
658
- | Step 3.5 | 3 core + flag-based conditionals | Full QCSD swarm with all 9 agents + DDD mapping |
659
- | Step 6 | SPARC-GOAP goal state analysis | Full agent with milestone tracking + success metrics |
659
+ | Step 3 | Pinned shift-left protocol | Same protocol; installed aqe tooling when available |
660
+ | Step 3.5 | Bundled QCSD protocol and conditional perspectives | Installed aqe runtime can coordinate its full swarm |
661
+ | Step 6 | SPARC-GOAP goal state analysis | Installed planner agent when available |
660
662
  | Step 7 | Standard code generation | Standard code generation (no change) |
661
- | Step 8 | Brutal-honesty 3-mode review | Full calibration levels (1-3) + evidence protocol + QE pattern recall/store |
662
- | Step 9 | 4 agents with condensed protocols | 4 agents with full agentic-qe protocols + fleet pattern recall/store |
663
+ | Step 8 | Pinned brutal-honesty protocol | Same review plus aqe pattern recall/store |
664
+ | Step 9 | 4 agents with pinned full protocols | Same agents plus aqe fleet pattern recall/store |
663
665
 
664
666
  ### What `--full-qe-extended` adds on top
665
667
 
@@ -227,7 +227,7 @@ Fitness readiness:
227
227
 
228
228
  After all ADRs are drafted, apply **shift-left-testing** protocol (Level 4: Risk Analysis in Design).
229
229
 
230
- > Load: `references/agentic-qe/shift-left-testing.md`
230
+ > Load: `references/agentic-qe/qe-ref-shift-left-testing/REFERENCE.md`
231
231
 
232
232
  ### Testability Check per ADR
233
233
 
@@ -6,7 +6,7 @@
6
6
 
7
7
  Adapted from [agentic-qe qcsd-ideation-swarm](https://github.com/proffesor-for-testing/agentic-qe) (Tier 3: Verified).
8
8
 
9
- > Load: `references/agentic-qe/qcsd-ideation-swarm.md`
9
+ > Load: `references/agentic-qe/qe-ref-qcsd-ideation-swarm/REFERENCE.md`
10
10
 
11
11
  ## When
12
12
 
@@ -13,7 +13,7 @@ Always runs, and ALWAYS writes `features/<slug>/06_implementation_plan.md` — e
13
13
 
14
14
  ## Model
15
15
 
16
- sonnet (analytical decomposition) + code-goal-planner SPARC-GOAP methodology
16
+ Use the planning model selected by the active feature-ADR routing policy, distinct from the coding model, with code-goal-planner SPARC-GOAP methodology.
17
17
 
18
18
  ## Input
19
19
 
@@ -25,7 +25,7 @@ sonnet (analytical decomposition) + code-goal-planner SPARC-GOAP methodology
25
25
 
26
26
  ## Protocol
27
27
 
28
- > Load: `references/agentic-qe/code-goal-planner.md` for SPARC-GOAP methodology.
28
+ In Claude Code, if the `code-goal-planner` agent is available, delegate goal-state analysis to it with the requirements, ADR decisions and quality risks. Check its proposed milestones against the project evidence before writing the plan. In Codex, or whenever that agent surface is unavailable, load `references/agentic-qe/code-goal-planner.md` and apply the same methodology directly. Do not report agent execution when only the reference was read.
29
29
 
30
30
  ### 0. Apply SPARC-GOAP Goal State Analysis
31
31
 
@@ -249,7 +249,7 @@ produced blockers. Non-blocking: empty result or error → proceed with the stan
249
249
 
250
250
  ### 3. Code Review via Brutal Honesty Review (M+)
251
251
 
252
- > Load: `references/agentic-qe/brutal-honesty-review.md`
252
+ > Load: `references/agentic-qe/qe-ref-brutal-honesty-review/REFERENCE.md`
253
253
 
254
254
  Apply the **brutal-honesty-review** protocol from agentic-qe (Tier 2: Validated).
255
255
 
@@ -28,11 +28,11 @@ sonnet (analytical fleet coordination — multiple parallel agents)
28
28
  ### 1. Fleet Assembly
29
29
 
30
30
  > Load agent protocols:
31
- > - Agent 1: `references/agentic-qe/qe-requirements-validation.md`
32
- > - Agent 2: `references/agentic-qe/risk-based-testing.md`
33
- > - Agent 3: `references/agentic-qe/enterprise-integration-testing.md`
34
- > - Agent 4: `references/agentic-qe/regression-testing.md`
35
- > - Shared: `references/agentic-qe/qe-coverage-analysis.md`
31
+ > - Agent 1: `references/agentic-qe/qe-ref-qe-requirements-validation/REFERENCE.md`
32
+ > - Agent 2: `references/agentic-qe/qe-ref-risk-based-testing/REFERENCE.md`
33
+ > - Agent 3: `references/agentic-qe/qe-ref-enterprise-integration-testing/REFERENCE.md`
34
+ > - Agent 4: `references/agentic-qe/qe-ref-regression-testing/REFERENCE.md`
35
+ > - Shared: `references/agentic-qe/qe-ref-qe-coverage-analysis/REFERENCE.md`
36
36
 
37
37
  Spawn a QE fleet of 4 parallel assessment agents:
38
38
 
@@ -100,7 +100,7 @@ Any UNMITIGATED critical risk (score ≥ 20) = GAP.
100
100
 
101
101
  ### 4. Integration Check (Agent 3)
102
102
 
103
- > Protocol from: `enterprise-integration-testing` (Tier 3, condensed)
103
+ > Protocol from: `enterprise-integration-testing` (pinned upstream protocol)
104
104
 
105
105
  Apply the **enterprise-integration-testing** hierarchical testing strategy:
106
106
 
@@ -251,17 +251,17 @@ Gaps requiring attention:
251
251
  ## Direct Mode (`--full-qe`)
252
252
 
253
253
  When `{AGENTIC_QE_MODE}` = `direct` (activated via `--full-qe` flag), Step 9
254
- uses full agentic-qe protocols for the same 4 core agents. No additional agents
255
- are added — the same fleet runs with deeper methodology from the full skill files.
254
+ uses the same four pinned full protocols as Reference Mode. No additional agents
255
+ are added; the installed aqe runtime adds its memory and fleet services.
256
256
 
257
257
  ### What changes
258
258
 
259
259
  | Agent | Reference Mode | Direct Mode (`--full-qe`) |
260
260
  |-------|---------------|---------------------------|
261
- | Agent 1 | Condensed qe-requirements-validation | Full SMART + BDD + sprint coverage protocol |
262
- | Agent 2 | Condensed risk-based-testing | Full 5×5 matrix + ML-enhanced + dynamic reassessment |
263
- | Agent 3 | Condensed enterprise-integration | Full hierarchical fleet + QCSD flags |
264
- | Agent 4 | Condensed regression-testing | Full change-based + historical + time-budget strategies |
261
+ | Agent 1 | Pinned qe-requirements-validation protocol | Same protocol; aqe runtime services |
262
+ | Agent 2 | Pinned risk-based-testing protocol | Same protocol; aqe runtime services |
263
+ | Agent 3 | Pinned enterprise-integration protocol | Same protocol; aqe fleet services |
264
+ | Agent 4 | Pinned regression-testing protocol | Same protocol; aqe runtime services |
265
265
 
266
266
  ## Pattern Memory Loop (Direct Modes)
267
267
 
@@ -0,0 +1,21 @@
1
+ MIT License
2
+
3
+ Copyright (c) 2025 Agentic QE Contributors
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
@@ -1,6 +1,6 @@
1
1
  # Agentic QE Skills Integration
2
2
 
3
- Skills from [agentic-qe](https://github.com/proffesor-for-testing/agentic-qe) integrated into the feature-adr pipeline.
3
+ Reference material from [agentic-qe](https://github.com/proffesor-for-testing/agentic-qe) integrated into the feature-ADR pipeline. Eight core skills are vendored from commit `0a6de058575cbe292afe220df105cc74120f7674` (MIT; see `LICENSE.agentic-qe`). The remaining reference files are Keysarium's older adapted copies.
4
4
 
5
5
  ## Installation (for Direct Mode)
6
6
 
@@ -51,7 +51,7 @@ Without installation, all core skills work in Reference Mode (no setup needed).
51
51
  ## Source Repository
52
52
 
53
53
  - Repository: https://github.com/proffesor-for-testing/agentic-qe
54
- - Skills path: `assets/skills/`
54
+ - Pinned skills path: `.claude/skills/`
55
55
  - Agents path: `.claude/agents/goal/`
56
56
  - Total skills in repo: 77+
57
57
  - Skills integrated: 15 (9 core + 6 extended)
@@ -79,13 +79,11 @@ Without installation, all core skills work in Reference Mode (no setup needed).
79
79
 
80
80
  ### Reference Mode (default, no setup)
81
81
 
82
- Skills are stored as condensed copies in `references/agentic-qe/`.
83
- Feature-adr modules read the skill instructions and apply their methodology
84
- within the feature-adr checkpoint framework.
82
+ The eight core skills listed below include their pinned upstream `SKILL.md`, output schemas, evals, and supporting files. Feature-ADR modules load those instructions from their own directories and apply the methodology within the feature-ADR checkpoint framework. Five skills also carry clearly marked Keysarium additions. The other reference files remain adapted copies.
85
83
 
86
84
  - Works out of the box
87
- - 15 files, ~50KB total
88
85
  - Covers all core pipeline steps
86
+ - No `aqe` executable, MCP tools, pattern memory, or fleet runtime is installed by these reference files
89
87
 
90
88
  ### Direct Mode (`--full-qe`, requires installation)
91
89
 
@@ -95,7 +93,7 @@ When `--full-qe` flag is used and agentic-qe is installed:
95
93
  2. Sets `{AGENTIC_QE_MODE}` = `direct`
96
94
  3. Modules load full skill files from agentic-qe package for the same 9 core skills
97
95
  4. Full QCSD swarm protocols with all agents + DDD mapping
98
- 5. No additional skills — same agents, deeper protocols
96
+ 5. The same pinned core protocols can use the installed runtime's CLI, MCP tools, memory and fleet agents when those capabilities are verified
99
97
 
100
98
  **When to use:**
101
99
  - L/XL tier features needing deeper QE methodology
@@ -123,16 +121,22 @@ Everything from `--full-qe` PLUS 6 additional skills:
123
121
 
124
122
  ## Files in this directory
125
123
 
126
- ### Core (always loaded)
127
- - `shift-left-testing.md` — Full copy (Tier 3)
128
- - `qcsd-ideation-swarm.md` — Condensed from ~77KB (Tier 3)
124
+ ### Core (loaded by the corresponding module)
125
+ - `qe-ref-shift-left-testing/REFERENCE.md` — pinned upstream plus Keysarium addition
126
+ - `qe-ref-qcsd-ideation-swarm/REFERENCE.md` — pinned upstream plus Keysarium addition
129
127
  - `code-goal-planner.md` — Full copy (Agent)
130
- - `brutal-honesty-review.md` — Full copy (Tier 2)
131
- - `qe-requirements-validation.md` — Full copy, 254 lines (Tier 3)
132
- - `risk-based-testing.md` — Full copy, 212 lines (Tier 3)
133
- - `enterprise-integration-testing.md` — Condensed from 735 lines (Tier 3)
134
- - `regression-testing.md` — Full copy, 233 lines (Tier 3)
135
- - `qe-coverage-analysis.md` — Full copy, 193 lines (Tier 3)
128
+ - `qe-ref-brutal-honesty-review/REFERENCE.md` — pinned upstream
129
+ - `qe-ref-qe-requirements-validation/REFERENCE.md` — pinned upstream
130
+ - `qe-ref-risk-based-testing/REFERENCE.md` — pinned upstream plus Keysarium addition
131
+ - `qe-ref-enterprise-integration-testing/REFERENCE.md` — pinned upstream plus Keysarium addition
132
+ - `qe-ref-regression-testing/REFERENCE.md` — pinned upstream plus Keysarium addition
133
+ - `qe-ref-qe-coverage-analysis/REFERENCE.md` — pinned upstream
134
+
135
+ Each vendored skill keeps its auxiliary files under the same directory, so its frontmatter paths such as `schemas/output.json` resolve from the skill root. `run-history.json` is excluded because it records upstream run history rather than the skill contract.
136
+
137
+ The `qe-ref-` directory prefix marks these as bundled references and keeps them distinct from separately installed skills with the same names. `UPSTREAM-MANIFEST.json` maps each local path to its original upstream path and hash.
138
+
139
+ The upstream `brutal-honesty-review/scripts/*.sh` files are bundled as reference utilities and are not run by Keysarium. They use simple text heuristics and, in one case, invoke the consumer project's `npm test`; their output is not a feature-ADR gate verdict. `assess-tests.sh` has a known upstream early-exit at its first `((found_count++))` under `set -e`; keep its verdict out of the gate. The JSON schemas are available for machine checks, but the pipeline does not yet claim that it runs every upstream eval automatically.
136
140
 
137
141
  ### Extended (loaded in Direct Mode)
138
142
  - `chaos-engineering-resilience.md` — Full copy, 164 lines (Tier 3)