@zrg-sh/studio 0.1.0

package/template/product-specs/workflows/pipeline-full.md

@@ -0,0 +1,258 @@
+# Pipeline: Full Feature
+
+2-layer pipeline with cross-reviews. Change package = business docs only.
+Executor writes code after change is done. Merger updates domain docs.
+
+## Naming
+
+- **PRDCT-NNNN-feature-name** -- change package directory name
+- Location: `product-specs/docs/changes/PRDCT-NNNN-feature-name/`
+
+## Roles
+
+| Alias | Agent | Notes |
+|-------|-------|-------|
+| PM | studio-pm | Product intent, business goals |
+| Scenario Writer | studio-scenario-writer | Gherkin scenarios |
+| Analyst | studio-analyst | System analysis |
+| Domain Framer | studio-domain-framer | Domain modeling |
+| Designer | studio-designer | UX flows, mockups |
+| Executor | studio-executor | Code implementation (backend + frontend) |
+| Merger | studio-domain-merger | Updates domain docs from change |
+
+## Pipeline overview
+
+```
+LAYER 1: WHAT & WHY
+  Stage 1    PM                    --> 01-intent.md
+  Stage 1.5  Analyst               --> cross-review-intent.md
+  Stage 2    Scenario Writer       --> 02-scenarios.md
+  Stage 2.5  Analyst               --> cross-review-scenarios.md (technically possible?)
+
+LAYER 2: HOW IT WORKS
+  Stage 3    Analyst               --> 03-analysis.md
+  Stage 3.5  PM                    --> cross-review-analysis.md (matches intent?)
+  Stage 4    Domain Framer + PM    --> 04-domain.md
+  Stage 4.5  Analyst               --> cross-review-domain.md (covers all scenarios?)
+
+LAYER 3: HOW IT LOOKS
+  Stage 5    Designer              --> 05-ux.md + mockups/
+  Stage 5.5  QA                    --> cross-review-ux.md (all scenarios covered?)
+
+POST-CHANGE
+  Executor                         --> backend + frontend code
+  Merger                           --> domain docs updated
+```
+
+---
+
+## Cross-review matrix
+
+```
+Stage  Author(s)            Reviewer(s)        Focus
+-----  --------             ----------         -----
+1.5    PM                   Analyst            Problem clear? Boundaries defined?
+2.5    Scenario Writer      Analyst            Scenarios technically possible? Edge cases?
+3.5    Analyst              PM                 System analysis matches business intent?
+4.5    Domain Framer + PM   Analyst            Domain model covers all scenarios? Invariants?
+5.5    Designer             QA                 Surfaces cover all scenarios? All states?
+```
+
+---
+
+## Gate rules
+
+Gates enforce ordering. A stage cannot start until its prerequisites pass.
+
+```
+Gate                          Requires
+----                          --------
+No scenarios --> no analysis  02-scenarios.md must exist before Stage 3
+No analysis --> no domain     03-analysis.md must exist before Stage 4
+No domain --> no UX           04-domain.md must exist before Stage 5
+No scenarios --> no UX        02-scenarios.md must exist before Stage 5
+Canon not edited directly     Domain docs updated only through merger
+```
+
+If a gate is not met, the pipeline stops and reports which artifact is missing.
+
+---
+
+## Execution
+
+### LAYER 1: WHAT & WHY
+
+#### Stage 1: Intent (PM)
+1. Run Agent: studio-pm.md + "Feature: $ARGUMENTS"
+2. Output: `01-intent.md` -- problem, goal, KPI, actors, boundaries
+3. STATE.md: `current_stage: 1`
+
+#### Stage 1.5: Cross-review -- Analyst reviews Intent
+Analyst reads `01-intent.md` and checks:
+- Problem clearly stated?
+- Boundaries well-defined?
+- KPI measurable?
+
+Output: `cross-review-intent.md`
+If NEEDS WORK --> return to PM (max 2 retries).
+
+---
+
+#### Stage 2: Scenarios (Scenario Writer)
+1. Run Agent: studio-scenario-writer.md + "PRDCT: $PRDCT_ID"
+2. Output: `02-scenarios.md` -- happy paths, edge cases, error scenarios, permissions, concurrent
+3. STATE.md: `current_stage: 2`
+
+#### Stage 2.5: Cross-review -- Analyst reviews Scenarios
+Analyst reads `02-scenarios.md` and checks:
+- Each scenario technically feasible?
+- Missing constraints? (rate limits, concurrency, file sizes)
+- Edge cases sufficient (>= 5)?
+
+Output: `cross-review-scenarios.md`
+If NEEDS WORK --> return to Stage 2 (max 2 retries).
+
+---
+
+### LAYER 2: HOW IT WORKS
+
+#### Stage 3: System Analysis (Analyst)
+**Gate: `02-scenarios.md` must exist.**
+
+1. Run Agent: studio-analyst.md + "PRDCT: $PRDCT_ID"
+2. Output: `03-analysis.md` -- system impact, technical constraints, integration points
+3. STATE.md: `current_stage: 3`
+
+#### Stage 3.5: Cross-review -- PM reviews Analysis
+PM re-reads `03-analysis.md` and checks:
+- System analysis matches business intent?
+- Constraints reasonable?
+- No scope creep?
+
+Output: `cross-review-analysis.md`
+If NEEDS WORK --> return to Stage 3 (max 2 retries).
+
+#### Stage 4: Domain (Domain Framer + PM)
+**Gate: `03-analysis.md` must exist.**
+
+1. Run Agent: studio-domain-framer.md + "PRDCT: $PRDCT_ID"
+2. PM provides business context, Domain Framer models the domain
+3. Output: `04-domain.md` -- aggregates, rules, events, invariants, ubiquitous language
+4. STATE.md: `current_stage: 4`
+
+#### Stage 4.5: Cross-review -- Analyst reviews Domain
+Analyst reads `04-domain.md` and checks:
+- Domain model covers all scenarios?
+- Terminology clear for non-technical stakeholders?
+- Invariants correct? Boundaries reasonable?
+
+Output: `cross-review-domain.md`
+If NEEDS WORK --> return to Stage 4 (max 2 retries).
+
+---
+
+### LAYER 3: HOW IT LOOKS
+
+#### Stage 5: UX (Designer)
+**Gate: `04-domain.md` AND `02-scenarios.md` must exist.**
+
+1. Run Agent: studio-designer.md + "PRDCT: $PRDCT_ID"
+2. Output: `05-ux.md` + `mockups/` directory with HTML files
+3. STATE.md: `current_stage: 5`
+
+#### Stage 5.5: Cross-review -- QA reviews UX
+QA reads `05-ux.md` and checks:
+- Every scenario from 02-scenarios.md has a corresponding surface?
+- All states represented (loading, success, error, empty)?
+- Error messages defined?
+- Permissions per surface specified?
+
+Output: `cross-review-ux.md`
+If NEEDS WORK --> return to Stage 5 (max 2 retries).
+
+**After Stage 5.5 passes: status -> done. Change package is complete.**
+
+---
+
+### POST-CHANGE FLOW
+
+#### Code Implementation (Executor)
+1. Run Agent: studio-executor.md + "PRDCT: $PRDCT_ID"
+2. Executor reads domain docs + scenarios + UX -> writes backend + frontend code
+3. Code derives API contracts, DB schema, components from domain knowledge + scenarios
+
+#### Domain Docs Update (Merger)
+1. Run Agent: studio-domain-merger.md + "PRDCT: $PRDCT_ID"
+2. Merger reads change package (01-05) -> updates domain docs:
+   - aggregates.md, business-rules.md, events.md, invariants.md
+   - ubiquitous-language.md, scenarios.md, surfaces.md
+   - api-contracts.md, data-model.md (from actual code, not docs)
+
+---
+
+## Cross-review format
+
+Every cross-review file follows this structure:
+
+```markdown
+# Cross-review: [Artifact] reviewed by [Reviewer]
+
+## Verdict: APPROVED / NEEDS WORK / BLOCKED
+
+## Findings
+| # | Finding | Severity | File:Section | Recommendation |
+|---|---------|----------|-------------|---------------|
+
+## Critical (must fix before next stage)
+-
+
+## Recommendations (nice to have)
+-
+```
+
+---
+
+## Iteration rules
+
+When a cross-review returns NEEDS WORK:
+
+1. Author revises the document based on findings
+2. Reviewer re-reads the revised document
+3. **Max 2 retries per cross-review stage**
+4. After 2 failed retries:
+   - Log the unresolved findings
+   - Proceed to next stage with a warning annotation
+   - STATE.md: `warnings: ["unresolved cross-review at stage X.5"]`
+
+When a cross-review returns BLOCKED:
+
+1. Record with reason
+2. Pipeline halts at that stage
+3. Requires human intervention to unblock
+
+---
+
+## Autonomous mode
+
+When invoked with `--auto`:
+
+- All cross-reviews execute automatically (no human prompts)
+- NEEDS WORK --> automatic revision by author (max 2 iterations per stage)
+- BLOCKED --> halt and notify
+- Skip all confirmation prompts
+- Each agent runs with full fresh context
+- Progress reported via STATE.md updates
+
+---
+
+## Final summary
+
+```
+PRDCT-NNNN-feature-name
+  Layer 1 (What & Why):  01-intent.md, 02-scenarios.md
+  Layer 2 (How it works): 03-analysis.md, 04-domain.md
+  Layer 3 (How it looks): 05-ux.md + mockups/
+  Cross-reviews:         N files
+  Warnings:              [any unresolved items]
+  Status:                DONE / IN PROGRESS / BLOCKED
+```

package/template/product-specs/workflows/pipeline-resume.md

@@ -0,0 +1,21 @@
+# Pipeline: Resume
+
+Возобновление работы с места остановки.
+
+## Process
+1. Прочитай `.planning/STATE.md` → определи current_chg, current_phase
+2. Если нет STATE.md → "Нет активной работы. Запусти /studio-pm для начала."
+3. Прочитай `product-specs/docs/changes/$PRDCT/metadata.yaml` → определи status
+4. Определи следующую стадию по stage-pipeline:
+   draft → /studio-analyst
+   discovery → /studio-analyst (continue)
+   spec → /studio-analyst (domain impact)
+   analysis → /studio-designer
+   design → /studio-backend
+   in-progress → определи что не заполнено (06? 07? 08?)
+   done → "PRDCT завершён."
+
+5. Покажи:
+   "Текущая позиция: PRDCT-XXXX, стадия: [X]
+    Следующий шаг: /studio-[role] PRDCT-XXXX"
+6. Спроси: "Запустить?"

package/template/product-specs/workflows/verify-change.md

@@ -0,0 +1,12 @@
+# Workflow: Verify Change
+
+## Input
+PRDCT-XXXX
+
+## Process
+1. Read entire change package
+2. Spawn studio-verifier agent with PRDCT context
+3. Receive VERIFICATION.md
+4. If verdict = GAPS → show gaps, ask user what to do
+5. If verdict = PASS → "PRDCT-XXXX verified. Ready for PR."
+6. If verdict = FAIL → show critical issues, recommend fixes