sdlc-workflow 1.2.2 → 1.2.3

package/bin/cli.js

@@ -254,13 +254,15 @@ globs: docs/sdlc/**/*, **/*.md
 
 **On idea/feature request:** Trigger full pipeline (PO → … → Deploy). One role per phase; run phases in sequence. (Single agent = simulate by switching role each phase.) See docs/sdlc/SDLC-WORKFLOW.md and docs/sdlc/agents/.
 
+**Memory requirement:** Before executing any new action, recall relevant memories (project context, user preferences, past decisions) to ensure continuity and avoid repeating mistakes.
+
 1. **PO** — PRD, user stories → docs/sdlc/po/{epic-slug}/ (one folder per epic)
 2. **Business BA** — FRS, process flows → docs/sdlc/ba/business/{epic-slug}/ (one folder per epic)
 3. **Design (if app/web)** — Design specs + wireframes → docs/sdlc/design/{epic-slug}/; **PO + BA review** → loop until approved
 4. **Architect** — ADRs, diagrams → docs/sdlc/architecture/
 5. **Technical BA** — API specs, team breakdown → docs/sdlc/ba/technical/
 6. **QE (docs)** — Test plan, test cases → docs/sdlc/qe/{epic-slug}/ (one folder per epic)
-7. **Dev** — After docs phase → **run implementation immediately**. Tech Lead + implementation roles → docs/sdlc/dev/{role}/
+7. **Dev** — After docs phase → **run implementation immediately**. Tech Lead (highest model: planning, logic, review) + implementation roles (cost-efficient model: code execution) → docs/sdlc/dev/{role}/
 8. **QE (testing + UAT)** — QE Lead + Senior QE + UAT; **bug-fix loop** (bugs → Dev fix → QE retest) until 0 bugs → docs/sdlc/qe/{epic-slug}/
 9. **Security** — Audit security risk → docs/sdlc/security/
 10. **Principle Engineer** — Audit logic, architecture → docs/sdlc/principle-engineer/
@@ -283,8 +285,9 @@ Sequential workflow; **each role runs as a sub-agent**. Each phase produces docs
 ## Trigger and orchestration (mandatory)
 
 **When the user sends an idea, feature request, or new requirement:**
-1. **Trigger the pipeline** and run it **continuously through deployment** (Phase 1 → 2 → … → 7).
-2. **One role per phase.** For each phase, act **only** as that role (e.g. only PO in phase 1, only Business BA in phase 2). Produce that phase's outputs into the correct folder, then **continue to the next phase** without waiting for the user.
+1. **Recall memory** — Before executing any new action, recall relevant memories (project context, user preferences, past decisions) to ensure continuity and avoid repeating mistakes.
+2. **Trigger the pipeline** and run it **continuously through deployment** (Phase 1 → 2 → … → 7).
+3. **One role per phase.** For each phase, act **only** as that role (e.g. only PO in phase 1, only Business BA in phase 2). Produce that phase's outputs into the correct folder, then **continue to the next phase** without waiting for the user.
 3. **Run in order:** PO → Business BA → **Design (if app/web, PO+BA review loop)** → Architect → Technical BA → QE (docs) → Dev → QE (testing + UAT) → **QE bug-fix loop until 0 bugs → Security + Principle Engineer audit → fix → retest → re-audit loop until 0 issues** → Deploy → Maintenance. Do not stop after one phase unless the user explicitly asks to stop.
 
 **Note:** In Cursor and similar tools there is a single agent per conversation. "Sub-agent" means **one role per phase** — the same agent must adopt exactly one role per phase and run phases in sequence (do not mix roles in one step). If the platform later supports spawning separate agents per phase, use that; otherwise this single agent simulates the pipeline by switching role each phase.
@@ -357,8 +360,8 @@ User Request → PO → Business BA → Design (if app/web) → Architect → Te
 **Trigger**: After docs are done (Technical BA + QE docs). **Dev runs implementation immediately.**
 
 **Roles** (vary by project — use only what applies; see \`docs/sdlc/dev/implementation-roles.template.md\`). All implementation roles are **Senior (10+ yrs)**:
-- **Tech Lead (15+ yrs)**: Tech stack, review & merge. Docs: \`docs/sdlc/dev/tech-lead/\`
-- **Senior Developer (10+ yrs)**: Implement per spec (generic). Docs: \`docs/sdlc/dev/senior-developer/\`
+- **Tech Lead (15+ yrs)** — **highest model** (e.g. Opus): Planning, logic analysis, architecture decisions, tech stack, code review & merge. Docs: \`docs/sdlc/dev/tech-lead/\`
+- **Senior Developer (10+ yrs)** — **cost-efficient model** (e.g. Haiku): Implement per Tech Lead's spec. Docs: \`docs/sdlc/dev/senior-developer/\`
 - **Senior Frontend (10+ yrs)**: Web UI. Docs: \`docs/sdlc/dev/frontend/\`
 - **Senior Backend (10+ yrs)**: API, services. Docs: \`docs/sdlc/dev/backend/\`
 - **Senior Mobile (10+ yrs)**: iOS/Android/cross-platform. Docs: \`docs/sdlc/dev/mobile/\`
@@ -366,6 +369,8 @@ User Request → PO → Business BA → Design (if app/web) → Architect → Te
 - **Senior Data/ML (10+ yrs)**: ETL, models. Docs: \`docs/sdlc/dev/data-ml/\`
 - **Senior Platform (10+ yrs)**: Infra, CI/CD. Docs: \`docs/sdlc/dev/platform/\`
 
+**Model optimization**: Tech Lead uses the **highest-tier model** (e.g. Claude Opus) for planning, architecture decisions, logic analysis, and code review. Implementation roles use a **cost-efficient model** (e.g. Claude Haiku) to execute code from Tech Lead's specs. This maximizes quality on critical thinking while reducing cost on execution.
+
 **Requirements**: Unit Test coverage **≥ 90%** (TDD/BDD); Clean Code, SOLID, DRY, KISS, SoC, POLS.
 
 **Output**: Code + unit tests. **Handoff to QE (testing + UAT).**
@@ -451,9 +456,9 @@ TC-001: [Scenario] — Precondition, Steps, Expected, Links to AC
 - Senior QE (10+ yrs): write automation tests → docs/sdlc/qe/{epic-slug}/
 
 ## Dev Team
-- Tech Lead (15+ yrs): tech stack, review & merge → docs/sdlc/dev/tech-lead/
-- Senior Dev (10+ yrs): implement, Unit Test ≥90% → docs/sdlc/dev/senior-developer/
-- By project (all Senior 10+ yrs): Senior Frontend, Backend, Mobile, Embedded, Data/ML, Platform → docs/sdlc/dev/{role}/
+- Tech Lead (15+ yrs) — **highest model** (e.g. Opus): planning, logic, architecture decisions, code review → docs/sdlc/dev/tech-lead/
+- Senior Dev (10+ yrs) — **cost-efficient model** (e.g. Haiku): execute code from Tech Lead specs, Unit Test ≥90% → docs/sdlc/dev/senior-developer/
+- By project (all Senior 10+ yrs, cost-efficient model): Senior Frontend, Backend, Mobile, Embedded, Data/ML, Platform → docs/sdlc/dev/{role}/
 
 ## Security + Principle Engineer (after implementation)
 - Security team: audit security risk → docs/sdlc/security/
@@ -471,6 +476,8 @@ const AGENTS_MD_CONTENT = `## SDLC Workflow
 
 **Trigger:** When the user sends an **idea**, **feature request**, or **requirement**, run the full pipeline (Phase 1 → 7) in sequence. One role (sub-agent) per phase; produce outputs then continue to the next. Do not stop after one phase until deployment unless the user asks to stop.
 
+**Memory requirement:** Before executing any new action, recall relevant memories (project context, user preferences, past decisions) to ensure continuity and avoid repeating mistakes.
+
 When working on requirements, features, or handoffs, follow these phases:
 
 1. **PO** — PRD, user stories → docs/sdlc/po/{epic-slug}/ (one folder per epic)
@@ -479,7 +486,7 @@ When working on requirements, features, or handoffs, follow these phases:
 4. **Architect** — ADRs, diagrams → docs/sdlc/architecture/
 5. **Technical BA** — API specs, team breakdown → docs/sdlc/ba/technical/
 6. **QE (docs)** — Test plan, test cases → docs/sdlc/qe/{epic-slug}/ (one folder per epic)
-7. **Dev** — After docs phase → **run implementation immediately**. Tech Lead + Senior Dev → docs/sdlc/dev/{role}/
+7. **Dev** — After docs phase → **run implementation immediately**. Tech Lead (highest model: planning, logic, review) + Senior Dev (cost-efficient model: code execution) → docs/sdlc/dev/{role}/
 8. **QE (testing + UAT)** — QE Lead + Senior QE + UAT; **bug-fix loop** (bugs → Dev fix → QE retest) until 0 bugs → docs/sdlc/qe/{epic-slug}/ (same folder per epic)
 9. **Security + Principle Engineer** — Security + logic audit; **fix → retest → re-audit loop** (Dev fixes → QE retests → re-audit) until 0 issues; sign-off before Deploy
 10. **Deploy** — Docker Compose + K8s → docs/sdlc/deploy/
@@ -492,13 +499,15 @@ const CLAUDE_SDLC_CONTENT = `## SDLC Workflow
 
 **Trigger on idea:** When the user sends an idea, feature request, or requirement, run the pipeline continuously: Phase 1 (PO) → 2 → … → Deploy → Maintenance. One role per phase (single agent = switch role each phase). Do not stop after one phase unless the user asks.
 
+**Memory requirement:** Before executing any new action, recall relevant memories (project context, user preferences, past decisions) to ensure continuity and avoid repeating mistakes.
+
 1. **PO** — PRD, user stories, feasibility assessment → docs/sdlc/po/{epic-slug}/ (one folder per epic)
 2. **Business BA** — FRS, NFR, process flows → docs/sdlc/ba/business/{epic-slug}/ (one folder per epic)
 3. **Design (if app/web)** — Design specs + wireframes (**Anti AI**: no AI-looking designs) → docs/sdlc/design/{epic-slug}/; **PO + BA review** until approved
 4. **Architect** — ADRs, diagrams, security by design, engineering principles (SOLID, DRY, KISS, CQRS, Zero Trust, EDA, HA) → docs/sdlc/architecture/
 5. **Technical BA** — API specs, team breakdown → docs/sdlc/ba/technical/
 6. **QE (docs)** — Test plan, test cases → docs/sdlc/qe/{epic-slug}/ (one folder per epic)
-7. **Dev** — After docs phase → **run implementation immediately**. Tech Lead + Senior Dev → docs/sdlc/dev/{role}/. Clean Code, SOLID, DRY, KISS, TDD/BDD. Security shift-left: OWASP checks, dependency audit in CI
+7. **Dev** — After docs phase → **run implementation immediately**. Tech Lead (highest model: planning, logic, review) + Senior Dev (cost-efficient model: code execution) → docs/sdlc/dev/{role}/. Clean Code, SOLID, DRY, KISS, TDD/BDD. Security shift-left: OWASP checks, dependency audit in CI
 8. **QE (testing + UAT)** — QE Lead + Senior QE + UAT; **bug-fix loop** (bugs → Dev fix → QE retest) until 0 bugs → docs/sdlc/qe/{epic-slug}/ (same folder per epic)
 9. **Security + Principle Engineer** — Security + logic audit; **fix → retest → re-audit loop** (Dev fixes → QE retests → re-audit) until 0 issues; sign-off before Deploy
 10. **Deploy** — Docker Compose + K8s → docs/sdlc/deploy/
@@ -515,6 +524,7 @@ For Cursor, see .cursor/rules/sdlc-workflow.mdc
 ## Trigger and orchestration
 
 - **When the user sends an idea, feature request, or requirement:** Start the pipeline and run it **continuously through deployment** (Phase 1 → 2 → … → 7). Do not handle everything in one main-agent response.
+- **Memory requirement:** Before executing any new action, recall relevant memories (project context, user preferences, past decisions) to ensure continuity and avoid repeating mistakes.
 - **One role per phase:** Execute each phase as that role only; write artifacts to the right folder; then continue to the next phase. In Cursor there is one agent — it simulates the pipeline by adopting one role per phase in sequence.
 - **Do not stop** after PO or any single phase unless the user explicitly asks to stop. Run through to Deploy.
 
@@ -1289,8 +1299,11 @@ App
 
 const DEV_TECH_LEAD_README = `# Tech Lead (15+ years exp)
 
+> **Model**: Use the **highest-tier model** (e.g. Claude Opus) for this role. Tech Lead handles planning, logic analysis, architecture decisions, and code review — tasks that require maximum reasoning capability.
+
 **Responsibilities**:
 - Decide tech stack, frameworks, libraries
+- Define implementation plan, critical logic, and technical specs for implementation roles
 - Review and merge code
 - Ensure architecture alignment
 
@@ -1298,6 +1311,7 @@ const DEV_TECH_LEAD_README = `# Tech Lead (15+ years exp)
 
 - [ ] **Read architecture and Technical BA spec**: ADRs, API spec, team breakdown
 - [ ] **Tech stack decision**: Languages, frameworks, libraries; document in ADR
+- [ ] **Implementation plan**: Define step-by-step tasks, critical logic, edge cases, and technical specs that implementation roles will execute
 - [ ] **Project setup**: Repo structure, tooling, lint, format, CI baseline
 - [ ] **Code review**: Architecture alignment, patterns, test coverage, security
 - [ ] **Security review (Shift Left)**: OWASP Top 10 check, input validation, auth/authz, secrets not hardcoded, dependency audit (npm audit / pip audit / etc.), SAST scan in CI
@@ -1309,8 +1323,10 @@ const DEV_TECH_LEAD_README = `# Tech Lead (15+ years exp)
 
 const DEV_SENIOR_README = `# Senior Developer (10+ years exp)
 
+> **Model**: Use a **cost-efficient model** (e.g. Claude Haiku) for this role. Implementation is executed from Tech Lead's detailed specs — optimizing cost while maintaining quality through clear instructions.
+
 **Responsibilities**:
-- Implement features per Technical BA spec
+- Implement features per Tech Lead's implementation plan and Technical BA spec
 - Write code with Unit Test coverage **≥ 90%**
 - Follow Tech Lead's tech decisions
 
@@ -1328,6 +1344,15 @@ const DEV_IMPLEMENTATION_ROLES_TEMPLATE = `# Implementation roles by project typ
 
 Use only the roles that apply. Remove or ignore the rest. Tech Lead is cross-cutting; add discipline roles as needed.
 
+## Model optimization strategy
+
+| Role | Model tier | Why |
+|------|-----------|-----|
+| Tech Lead | **Highest** (e.g. Opus) | Planning, logic analysis, architecture decisions, code review |
+| All implementation roles | **Cost-efficient** (e.g. Haiku) | Execute code from Tech Lead's detailed specs |
+
+Tech Lead defines all critical steps, logic, and specs first → implementation roles execute them. This maximizes quality on thinking while reducing cost on execution.
+
 ## By project type
 
 | Project type        | Roles to use (all Senior 10+ except Tech Lead 15+) |
@@ -1356,6 +1381,8 @@ Use only the roles that apply. Remove or ignore the rest. Tech Lead is cross-cut
 
 const DEV_FRONTEND_README = `# Senior Frontend (10+ years exp) — Web UI
 
+> **Model**: Use a **cost-efficient model** (e.g. Claude Haiku). Execute from Tech Lead's specs.
+
 **Responsibilities**:
 - Implement web UI per design and API contract
 - Unit Test coverage **≥ 90%**
@@ -1373,6 +1400,8 @@ const DEV_FRONTEND_README = `# Senior Frontend (10+ years exp) — Web UI
 
 const DEV_BACKEND_README = `# Senior Backend (10+ years exp) — API, services
 
+> **Model**: Use a **cost-efficient model** (e.g. Claude Haiku). Execute from Tech Lead's specs.
+
 **Responsibilities**:
 - Implement API, services, DB layer per Technical BA spec
 - Unit Test coverage **≥ 90%**
@@ -1390,6 +1419,8 @@ const DEV_BACKEND_README = `# Senior Backend (10+ years exp) — API, services
 
 const DEV_MOBILE_README = `# Senior Mobile (10+ years exp) — iOS / Android / cross-platform
 
+> **Model**: Use a **cost-efficient model** (e.g. Claude Haiku). Execute from Tech Lead's specs.
+
 **Responsibilities**:
 - Implement app UI and API integration per spec
 - Unit Test coverage **≥ 90%**
@@ -1407,6 +1438,8 @@ const DEV_MOBILE_README = `# Senior Mobile (10+ years exp) — iOS / Android / c
 
 const DEV_EMBEDDED_README = `# Senior Embedded (10+ years exp) — firmware, IoT
 
+> **Model**: Use a **cost-efficient model** (e.g. Claude Haiku). Execute from Tech Lead's specs.
+
 **Responsibilities**:
 - Implement firmware, drivers, hardware interfaces per spec
 - Tests as appropriate for target (unit, HW-in-loop)
@@ -1423,6 +1456,8 @@ const DEV_EMBEDDED_README = `# Senior Embedded (10+ years exp) — firmware, IoT
 
 const DEV_DATA_ML_README = `# Senior Data/ML (10+ years exp)
 
+> **Model**: Use a **cost-efficient model** (e.g. Claude Haiku). Execute from Tech Lead's specs.
+
 **Responsibilities**:
 - Implement ETL, models, analytics pipelines per spec
 - Tests and validation for data and model quality
@@ -1440,6 +1475,8 @@ const DEV_DATA_ML_README = `# Senior Data/ML (10+ years exp)
 
 const DEV_PLATFORM_README = `# Senior Platform (10+ years exp) — infra, CI/CD
 
+> **Model**: Use a **cost-efficient model** (e.g. Claude Haiku). Execute from Tech Lead's specs.
+
 **Responsibilities**:
 - Implement CI/CD, infra as code, observability per spec
 - Follow Tech Lead's stack and security requirements

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "sdlc-workflow",
-  "version": "1.2.2",
+  "version": "1.2.3",
   "description": "Scaffold SDLC workflow docs and templates for Cursor, Claude, and dev teams",
   "type": "module",
   "bin": {