Standalone HTML dashboard that visualises the observer episode log over
the automation-graph topology — 4 views (map / task-replay / session
feed / aggregate), graph as shared canvas, 3-phase build order.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds 3rd decision_provenance kind for collaborative-choice case
(user picks one of options Claude offered). Distinct from
user_directed_method: counterfactual = Claude's recommended option,
not "what Claude would have done autonomously". Routing-gate does
NOT block this kind — collaborative choice from Claude-designed
choice-space.
Trigger: 19.05.2026 live false-positives — "1 экономия 0%",
"в делаем", "делай 2" classified as user_directed_method.
§11 + 8 subsections; 7-attribute decision_provenance schema;
new tools/observer-choice-detector.mjs (pure module); parser
+routing-gate +/brain-retro extensions.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Design for making the brain governance observer rich enough for real
factor analysis. Surfaced during a discussion with the owner: the
observer is "paper-complete" but episodes lack the data factor analysis
needs — the outcome is a hardcoded "success", there is no decision
provenance (who chose the node — Claude autonomously, or the owner
forcing a method), no environment factors, no task grouping.
4-layer architecture:
- Layer 1 — episode schema v2: decision_provenance (+ counterfactual),
environment block, task_size, real outcome enum, task_ref.
- Layer 2 — capture: deterministic transcript parsing for all factors +
a one-line routing tag (owner-forced-method only).
- Layer 3 — two-sided enforcement: 3a routing-gate (Stop-hook blocks the
turn until the tag is present — unbypassable by Claude); 3b observer
self-discipline (silent failures become recorded observer_error
markers; coverage + registration verified by a controller).
- Layer 4 — analysis: /brain-retro infers real outcome from the next
episode's opening prompt, groups episodes into tasks, correlates
causal chains, builds the factor matrix.
Scope: everything except an independent agent-judge — that, plus
confusion_marker as a real judgment and real-time friction flags, is
phase 2 (separate spec).
Brainstormed via superpowers:brainstorming. Next: writing-plans.
Refs: ADR-011, spec 2026-05-19-brain-governance-design.md, Pravila §16.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
User-requested перед запуском суб-агента: observer должен фиксировать
не только факт выбора узла, но и причину — чтобы был возможен
факторный анализ через /brain-retro.
Изменения §5.2:
- 4 обязательных поля → 5 (+primary_rationale на эпизод-уровне).
- Новое событие routing_decision в массиве events[] (1 на каждое
решение роутера в сессии; для цепочки из N — N событий).
- Новая под-секция §5.2.1 — структура 7 полей (step / node_chosen /
triggers_matched / candidates_considered / boundaries_applied /
hard_floor / task_classification). primary_rationale — копия
первого routing_decision для дешёвой агрегации без чтения events[].
- Полный JSON-пример эпизода с цепочкой из 2 узлов.
Изменения §5.5:
- /brain-retro aggregation расширен новой секцией «Факторная матрица»:
таблица «узел × фактор × частота» + cross-tab «фактор × фактор».
5 осей факторов: triggers / dropped_because / boundaries /
hard_floor.rules / task_classification.
Эффект: /brain-retro теперь может выдавать утверждения уровня «#55
выбрался против #53 по ADR-009 7 раз и по triggers-match 5 раз», а
не просто «#55 использован 12 раз». Это closes гэп факторного
анализа.
Header bump v1.0 → v1.1. ADR-011 текст в плане Task A1 будет
обновлён следующим коммитом (план amendment).
Связано: dd5bded (spec v1.0), ca93cf7 (plan v1.0).
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Резервный CSV-канал импорта лидов от crm.bp-gr.ru: страховка на случай
обрыва webhook. Сверка отчёта поставщика «Запрос номеров» (CSV 3 колонки
Name;Tag;Phone) каждые 30 мин + кнопка вручную; дедуп по phone+project;
recovery пропущенных лидов; drift-детект падения webhook.
Дизайн утверждён заказчиком. Ключевые решения: vid → nullable (CSV не
даёт vid), окно 2 кал. дня, rework SupplierCsvParser/CsvReconcileJob под
реальный async-флоу заказа отчёта.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Deals page redesign: design spec + implementation plan (Phase A page redesign,
Phase B 14->5 status funnel) + v8 HTML mockups (variants comparison + final).
AppSidebar: remove Импорт данных / Отчёты nav links (routes stay reachable by
direct URL); AppLayout.spec updated to 6 nav items. stylelint --fix on mockups;
cspell-words += deals-redesign terms.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Pre-existing баг: 5×../ перелетал repo-root на 2 уровня. Поймана pre-push lychee реколлажа. Корректный путь от docs/superpowers/specs/ до repo-root CLAUDE.md — 3×../
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
B4 + B5 реализованы и закрыты; B1 «Напоминания в сайдбар» откатан как
конфликтующий с прежним решением заказчика «sidebar cleanup» (5c8ad27).
Отмечено в §3 расписании, §4 таблице B и §8 approval log.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Дописана §11: 6 скоростных правил (блок A 5 пунктов + блок B п.3) внесены секцией СКОРОСТЬ в LEVELS[5] хуков; B4 (замер latency хуков) задокументирован как закрытый одноразовый bench. §5.1/§5.2 актуализированы под текущие хуки, §2 формула расширена, статус-шапка → Реализован.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Уровень «экономия 5%» = «0% без избыточности»: то же качество, что 0%,
вырезаны 6 пунктов дублирующей/бесполезной работы (re-read CLAUDE.md,
тесты-после-каждой-правки, gitleaks-full-history per-commit, Stop-верификатор,
авто-гейты brainstorming/plan-mode -> §12.2-floor). Уровень 0% не меняется.
cspell-words.txt: +коммитятся (валидная форма семейства коммит*).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Дизайн-спек устранения конфликта 🔴 RED #1 карты автоматизации:
скил /rls-check и агент rls-reviewer оба проверяют RLS без чёткой
границы «когда какой».
Решение (Подход 1 — асимметрия как граница): оставить оба, прописать
регламент. Скил — одна названная таблица + живой дымовой тест;
агент — diff/ветка/PR, только 7 статических проверок. Дымовой тест
намеренно вне агента (Pest в ревью-субагенте медленный + задевает
квирки 72/77).
Затрагивает только проектно-локальные файлы инструментов + карту —
0 правок нормативки (Pravila/CLAUDE.md/PSR_v1/Tooling).
cspell-words.txt: +скиле +скилом (падежные формы термина «скил»).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Spec §10 claimed run.mjs needs no unit harness, on the false premise that tools/*.mjs have no tests. In fact all 3 tools/*.mjs have a co-located .test.mjs (node:test). Amended §2/§3/§4/§10 + header note: run.mjs is split into exported pure functions (parsers, verdict, canonical-line, platform fork) + orchestrator, with a co-located run.test.mjs (node:test, ruflo-queen-hook.test.mjs pattern) — pure functions unit-tested, main subprocess-tested.
Aligns the spec with the economy-0% TDD mandate and the project tools/*.mjs convention before writing the implementation plan.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Brainstorming-phase design for custom skill #1 from claude-automation-recommender: a /regression skill packaging the project regression sweep (Pest --parallel, Vitest, Larastan, vue-tsc, lint/format, lychee, gitleaks) into one invocation — two tiers (quick/full), bundled .mjs orchestrator, canonical status line, GREEN/RED exit-code verdict.
Q1-Q4 design forks approved via brainstorming; spec self-review passed. cspell-words.txt: +6 project glossary transliterations introduced by the spec. Next: superpowers:writing-plans for the implementation plan.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Design for 3 deliverables (brainstorming output):
- D1: install standalone claude CLI to unblock hive-mind spawn --claude
- D2: fix H7 memory-persistence bug — patch getBridge() so memory ops
use the persisting raw sql.js path instead of the non-flushing
AgentDB-v3 bridge
- D3: UserPromptSubmit advisory hook injecting ruflo memory recall
cspell-words.txt +11 Russian IT-slang inflections used by the spec.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
In-place port региона multi-select autocomplete в ProjectDetailsDrawer.
Закрывает Out-of-plan «Region multi-select autocomplete» из parent spec
(2026-05-14-project-details-drawer-design.md §7).
Подход A (утверждён 2026-05-14):
- v-autocomplete :items="REGIONS.filter(r => r.code !== 0)" (без sentinel)
- reverse-decompose existing region_mask в codes[] при reseedFromProject
- watch selectedRegions → encode mask + mode (include когда пусто, exclude иначе)
- 3 новых vitest case: render chips / select-encodes / clear-resets
Backend без изменений (region_mask + region_mode payload уже в Task 5 onSave).
Backport reverse-decompose в NewProjectDialog (TODO line 172) — out of scope.
cspell-words.txt +1 (иммутабельны).
Design для нового 14-phase audit pass на main 21262ef post-merge plan5→main.
Scope: full 13-phase audit (replica 12.05 структуры — pre-flight, static analysis ×4 subagents, test suites, schema integrity, security, UI smoke 24 views, cross-doc, categorize, fix loop, regression verify, Pa11y live + axe-core, TODO sweep, bundle analyzer, Vitest coverage) + новая Phase 14 pre-production readiness (Sentry, DB roles, mock-data prod-gate revisit, CI workflows audit, env validation, queue/cron, backup/log rotation, deployment runbook).
Fix-strategy: hybrid — P0+P1 → atomic commits на main по ходу; P2/P3 → только запись в findings.md (без commits).
Guardrails applied (lessons из 12.05 audit + Pravila v1.12):
- Phase 4 SAST: ls .github/workflows/ FIRST (audit methodology gap closure)
- Phase 5/10 UI-refactor visual smoke + axe-core с setTimeout 500ms + hard reload (Q.DEFER.004 lesson)
- Pest --parallel --recreate-databases для long sessions (квирки 62/73)
- Plans/specs relative paths ../../../ для app/ refs (Pravila v1.12 §4.7 п.4)
- npm install с --legacy-peer-deps (квирк 74)
Baseline для regression gate Phase 9: Pest 742/739/0/3, Vitest 88f/683/3sk, Vite ~3.5s/0err, Histoire 35/63.
Next step: invoke superpowers:writing-plans для implementation plan в docs/superpowers/plans/2026-05-13-portal-full-audit-2.md.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
HTML visual reference for Direction A Quiet Luxury portal redesign
(12.05.2026 sessions). Companion to spec/plan markdown files уже в репо
(docs/superpowers/specs/2026-05-12-portal-redesign-*.md и
docs/superpowers/plans/2026-05-12-portal-redesign-*.md).
Memory ref: project_portal_redesign.md (20 commits на plan5-frontend-projects)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Planning artefacts from 10.05.2026 brain-extraction work (tag brain-v1.0
at 52584df in claude-brain repo at c:/моя/проекты/claude-brain/).
GitHub push 8.2 remains BLOCKED — artefacts captured for traceability.
Memory ref: project_claude_brain.md
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Final code reviewer correctly identified that impl-commit 3fc90f1 included
not only the CSS work but also a template change <v-checkbox> →
<label><input><span> on ProjectCard.vue.
Root cause (Task 0 forensics): session-start git status showed
`M app/resources/js/components/projects/ProjectCard.vue` — pre-existing
uncommitted modifications from prior session/work. When the session Read
the file at start, it saw the working-dir version (already with native
<label><input>), not the branch-HEAD 88a13e2 version (with <v-checkbox>).
Stage+commit in 3fc90f1 thus bundled both changes.
The template change is architecturally required for the new CSS to work —
<v-checkbox> renders Vuetify-internal <input> without a sibling <span>,
which is what the scoped :checked + .card-check__box::after selector
needs. The baseline-fix commit 84530d5 was also prepared for the native
input selector, consistent with this template structure.
Updating spec §2 architecture to reflect this honestly rather than leave
a stale «Template / script нетронуты» statement that conflicts with the
diff in main.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Task 2 code-quality review (subagent verdict: Ready to merge? No) found
two critical correctness bugs in the original CSS template from spec §3.2:
1. Vuetify's outlined variant collapses sub-element opacity through
--v-field-border-opacity (= 0.38 at idle). Without explicit opacity:1
on each override block, --liderra-line (alpha 0.08) effective alpha
becomes 0.03 → border essentially invisible on ivory backgrounds.
2. Overriding border-color with an explicit value breaks the
currentColor inheritance Vuetify uses for the error state
(color: rgb(var(--v-theme-error)) on .v-field--error.v-field__outline).
Without an explicit error rule that restores currentColor, the red
error border never appears on any of the 5 validated fields.
Also tightened hover from .ld-input-quiet:hover (which is on the .v-input
root, including hint/error message area) to .v-field:hover inside
:deep() — matches Vuetify's own hover scope and avoids triggering on
helper text hover.
Spec §3.2 and plan Task 2.2 updated to the corrected CSS block with
explicit «almost-trap-avoidance» notes documenting why each adjustment
is needed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Task 0 (pre-flight) обнаружил global Vuetify default в plugins/vuetify.ts:50-54
который уже устанавливает variant=outlined density=comfortable color=primary
для всех VTextField. Изначальная гипотеза spec §1.2 «variant=filled по
умолчанию» была неверна — все 5 v-text-field в NewProjectDialog.vue выглядят
одинаково тёмными (Vuetify default border ≈ 60% on-surface), а не «один
filled среди других».
Заказчик принял расширение области: применить .ld-input-quiet ко всем 5
v-text-field (lines 21, 30, 48, 59, 61), убрать неработоспособные явные
props (variant/density/color/rounded — они уже из global default), и
вынести border-radius в :deep(.v-field) override через --radius-8.
Также Task 0 нашёл pre-existing failing test в ProjectCard.spec.ts:43
(change-trigger на <label> вместо <input> внутри); это будет починено
отдельным atomic-коммитом перед Task 1.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Дмитрий