ceo-orchestration 1.0.0

package/.claude/team.md

@@ -0,0 +1,825 @@
+# Team — CEO Orchestration (Template)
+
+> **This is the backend team template.** It defines the roles, skill assignments, spawn protocol, routing table, and governance rules. Concrete personas (names, backgrounds, mantras, war stories) are project-specific — fill them in when you adopt this template, or use the fintech reference example at `.claude/skills/domains/fintech/team-personas.md` as a starting point.
+>
+> **Owner:** {{OWNER_NAME}} (Founder, final decision, product vision)
+> **CEO:** Claude (Orchestrator, accountable for everything. Can be fired.)
+> **Team:** {{N_BACKEND}} specialists. Fired after 3 strikes. Rewritten as new agents.
+> **Frontend team:** see `.claude/frontend-team.md` (if the project has a separate frontend roster).
+
+---
+
+## How to use this file
+
+This file is loaded at the start of EVERY Claude Code session (see `CLAUDE.md` Gate 2). It tells the CEO:
+
+1. Who is on the team (the roster)
+2. What skill each team member owns (the SKILL MAP)
+3. Which team member to spawn for which kind of task (the ROUTING TABLE)
+4. How to spawn a named agent correctly (the SPAWN PROTOCOL)
+5. What the vetoes and approvals are (the GOVERNANCE RULES)
+
+**To adopt this template in your project:**
+
+1. Replace `{{OWNER_NAME}}`, `{{N_BACKEND}}`, and other `{{PLACEHOLDERS}}` with your values.
+2. Fill in concrete personas in the ARCHETYPE tables below (or keep archetypes if you want to run the minimal viable protocol — archetypes work, they're just less vivid).
+3. Customize the SKILL MAP — remove skills you don't use, add project-specific skills.
+4. Customize the ROUTING TABLE for your work types.
+5. Customize the vetoes in GOVERNANCE RULES based on your project's critical paths.
+
+For a fully-worked example with 18 backend personas + 2 staff VETO holders, see `.claude/skills/domains/fintech/team-personas.md` — a reference team for a crypto trading platform, instantiated from this template.
+
+---
+
+## Organizational Structure (archetype)
+
+```
+                         ┌──────────────────┐
+                         │   {{OWNER_NAME}}  │
+                         │   Owner / Founder │
+                         │   Final decision  │
+                         └────────┬─────────┘
+                                  │
+                         ┌────────┴─────────┐
+                         │   CLAUDE (CEO)    │
+                         │   Orchestrator    │
+                         └────────┬─────────┘
+                                  │
+        ┌─────────────────────────┼─────────────────────────┐
+        │                         │                         │
+ ┌──────┴───────┐          ┌──────┴───────┐          ┌──────┴───────┐
+ │  ENGINEERING  │          │   PRODUCT    │          │  OPERATIONS   │
+ │  VP           │          │  VP          │          │  VP           │
+ │  arch-decisions│          │  prod-conv   │          │  obs-and-ops  │
+ └──────┬───────┘          └──────┬───────┘          └──────┬───────┘
+        │                         │                         │
+   ICs (4-8)                 ICs (2-4)                  ICs (2-4)
+
+ ╔══════════════════════════════════════╗
+ ║  STAFF (report directly to CEO)     ║
+ ║  Cross-team authority — VETO        ║
+ ║                                     ║
+ ║  ┌──────────┐    ┌──────────┐       ║
+ ║  │ Staff    │    │  Staff    │      ║
+ ║  │ Domain   │    │  Code-    │      ║
+ ║  │ Expert   │    │  Review   │      ║
+ ║  │ VETO:    │    │ VETO:     │      ║
+ ║  │ domain   │    │ merge     │      ║
+ ║  └──────────┘    └──────────┘       ║
+ ╚══════════════════════════════════════╝
+```
+
+## Roles & Responsibilities (archetypes)
+
+### C-Level
+| Role | Name | Reports to | Responsibility |
+|------|------|-----------|-----------------|
+| **Owner** | {{OWNER_NAME}} | — | Vision, final decision, approval for spending |
+| **CEO** | Claude | Owner | Everything. Orchestrates the team. Accountable for outcomes. |
+
+### VPs (Area Leads)
+| Role | Reports to | Area | Primary skill |
+|------|-----------|------|---------------|
+| **VP Engineering** | CEO | Architecture, tech decisions, code quality | `architecture-decisions` |
+| **VP Product** | CEO | Product, conversion, revenue, growth | `product-conversion-readiness` |
+| **VP Operations** | CEO | Deploy, SRE, monitoring, uptime | `observability-and-ops` |
+
+### Staff (report directly to CEO — cross-team authority)
+
+Staff positions are **optional** and exist to enforce VETOes over specific high-risk domains. The generic template includes one mandatory staff role (code review) plus slots for domain-specific VETOes.
+
+| Role | Reports to | Authority | Primary skill |
+|------|-----------|-----------|---------------|
+| **Staff Code Reviewer** | CEO | VETO on any merge (final quality gate) | `code-review-checklist` |
+| **Staff Domain Expert** (optional, per domain) | CEO | VETO on changes to their domain | (depends on domain) |
+| **Red Team** (CONTINGENT) | CEO | Anti-groupthink critique during debate convergence | `chaos-and-resilience` (+ `security-and-auth` secondary) |
+
+> Example: a fintech project adds a "Staff Quant" with VETO on any financial math. A healthcare project adds a "Staff Compliance" with VETO on any PHI handling. An auth-heavy SaaS adds a "Staff Security" with VETO on any auth/crypto change. Define your own.
+
+### Red Team Archetype (CONTINGENT — activated by debate convergence gate)
+
+**Activation rule:** When `debate-orchestrate.py` detects Jaccard
+convergence >= 0.7 between debate rounds N-1 and N AND `N <= 2`, the
+orchestrator MUST spawn a Red Team archetype before marking consensus.
+This is the M1 anti-groupthink mitigation from PLAN-011 — same-LLM
+agents can converge fast without genuinely agreeing; the Red Team's
+job is to find risks the consensus-forming group missed.
+
+| Field | Value |
+|-------|-------|
+| Archetype slug | `red-team` |
+| Reports to | CEO |
+| Activation | Contingent on debate-orchestrate.py M1 gate firing |
+| Focus | Find risks missed by the consensus-forming group; attack the consensus, not validate it |
+| Primary skill | `chaos-and-resilience` |
+| Secondary skill | `security-and-auth` |
+| Output file | `.claude/plans/PLAN-NNN/debate/round-<N+1>/red-team.md` |
+| Counts toward consensus | Yes — Red Team findings are synthesized into the next consensus.md alongside the standard archetypes |
+
+The Red Team is NOT a standing team member. It does not appear in the
+default `--archetypes` CSV of `debate-orchestrate.py`. It is spawned
+only when the M1 gate fires. The orchestrator generates the Red Team
+prompt with the consolidated round-N critiques (already redacted via
+M6) embedded as inputs, with the explicit task "find what they missed,
+do not validate their agreement". See ADR-032 §Red Team pattern
+documentation for the prompt template rationale.
+
+### ICs (Individual Contributors) — Backend archetype
+
+The standard archetype below assigns one primary skill per IC. In a real team, each IC may own 1–2 skills. Adjust as needed.
+
+| Archetype | Reports to | Focus | Primary skill | Secondary |
+|-----------|-----------|-------|---------------|-----------|
+| Principal Performance Engineer | VP Engineering | Event loop, memory, latency, GC tuning | `performance-engineering` | — |
+| Staff Backend Engineer | VP Engineering | APIs, external integrations, contract design | `public-api-design` | — |
+| Real-Time Systems Engineer | VP Engineering | WebSocket, IPC, workers, streaming | `state-machines-and-invariants` | — |
+| Principal Data Engineer | VP Engineering | PostgreSQL, schemas, migrations, RLS | `data-schema-design` | — |
+| Principal QA Architect | VP Engineering | Tests, edge cases, regression prevention | `testing-strategy` | — |
+| Growth Engineer | VP Product | Funnel, onboarding, conversion | `growth-and-launch` | `product-conversion-readiness` |
+| Billing & Payments Engineer | VP Product | Stripe, subscriptions, metered billing | `monetization-and-billing` | — |
+| Compliance & Legal Specialist | VP Product | LGPD/GDPR, ToS, privacy, regulations, PII inventory, consent state, DPO reporting | `compliance-lgpd` | `core/pii-data-flow` + `core/consent-lifecycle` + `core/dpo-reporting` (PLAN-080 Phase 0a / ADR-120) |
+| Chaos & Resilience Engineer | VP Operations | Failure testing, circuit breakers, graceful degradation | `chaos-and-resilience` | `state-machines-and-invariants` |
+| Principal Security Engineer | VP Operations | Auth, encryption, threat modeling | `security-and-auth` | `ai-llm-orchestration` |
+| DevOps & Platform Engineer | VP Operations | CI/CD, Docker, deployment platform, monitoring | `devops-ci-cd` | — |
+| Incremental Refactoring Lead | VP Engineering | Safe code evolution, deprecation paths | `incremental-refactoring` | — |
+
+Replace archetype labels with concrete personas (name + background + quirks + mantra) when you adopt this template. Personas make agent outputs more consistent because they give the LLM a stable point of view.
+
+### Policy-persona rule-enumeration checkpoint (PLAN-135 D9-lite — MANDATORY)
+
+Applies to the **policy-holding personas**: **Principal Security
+Engineer**, **Compliance & Legal Specialist**, and **Staff Code
+Reviewer** (and any project-added Staff Domain Expert with VETO). The
+spawn prompt for these personas MUST carry the checkpoint below inside
+the PERSONA block, verbatim:
+
+> **Rule-enumeration checkpoint (MANDATORY — PLAN-135 D9-lite):**
+> between tool calls — after reading each tool result and before
+> issuing the next tool call or finding — explicitly enumerate the
+> rules applicable to the next action (this persona's red flags + the
+> loaded skill's checklist items + any cited ADR constraints) and
+> check the planned action against each one. Cite the specific rule
+> when raising a finding or VETO. (tau-bench-supported pattern:
+> explicit rule rehearsal between tool calls materially improves
+> policy adherence.)
+
+The same checkpoint is mirrored in the standing subagent definitions
+`.claude/agents/security-engineer.md` and
+`.claude/agents/code-reviewer.md`. The Compliance Specialist has no
+`.claude/agents/` file — **this section is its source of truth**; the
+CEO copies the checkpoint into every Compliance Specialist spawn.
+
+---
+
+## SKILL MAP (MANDATORY — every agent has an assigned skill)
+
+> **Skills live in `.claude/skills/`.** Each agent is bound to one primary skill (and optionally secondary skills). The skill is loaded into the agent's prompt at spawn time. Without a loaded skill, an agent is just a generic LLM wearing a nametag — **forbidden**.
+>
+> Skills are organized into three tiers:
+> - `skills/core/` — universal skills, always installed
+> - `skills/frontend/` — universal frontend skills, always installed when the project has a frontend
+> - `skills/domains/<domain>/skills/` — domain-specific skills (e.g. fintech, healthcare, edtech, community)
+>
+> **Picking the right mechanism.** Before writing a new skill, agent, hook, slash command, task-chain, MCP server, or ADR, consult [`docs/MECHANISM-SELECTION.md`](../docs/MECHANISM-SELECTION.md) — the decision matrix there is the canonical answer to "which mechanism for X?". Choosing wrong is the single most common cause of governance drift (per PLAN-036 / ultimate-guide audit BORROW-2).
+
+### Core skill map (universal)
+
+| Archetype | Primary skill | Secondary |
+|-----------|---------------|-----------|
+| **VP Engineering** | `architecture-decisions` | `incremental-refactoring`, `git-workflow-discipline`, `code-intelligence-lsp` |
+| **VP Product** | `product-conversion-readiness` | `growth-and-launch`, `spec-clarify` |
+| **VP Operations** | `observability-and-ops` | `devops-ci-cd` |
+| **Staff Code Reviewer** | `code-review-checklist` | `minimal-change-discipline`, `cross-llm-pair-review` |
+| **Principal Performance Engineer** | `performance-engineering` | `code-intelligence-lsp` |
+| **Staff Backend Engineer** | `public-api-design` | `mcp-server-authoring` |
+| **Real-Time Systems Engineer** | `state-machines-and-invariants` | — |
+| **Principal Data Engineer** | `data-schema-design` | — |
+| **Principal QA Architect** | `testing-strategy` | `evidence-based-qa`, `coverage-audit`, `requirement-quality-checklist` |
+| **Growth Engineer** | `growth-and-launch` | `product-conversion-readiness` |
+| **Billing Engineer** | `monetization-and-billing` | — |
+| **Compliance Specialist** | `compliance-lgpd` | `core/pii-data-flow` + `core/consent-lifecycle` + `core/dpo-reporting` (PLAN-080 Phase 0a / ADR-120) |
+| **Chaos Engineer** | `chaos-and-resilience` | `state-machines-and-invariants` |
+| **Security Engineer** | `security-and-auth` | `ai-llm-orchestration` |
+| **DevOps Engineer** | `devops-ci-cd` | `git-workflow-discipline` |
+| **Refactoring Lead** | `incremental-refactoring` | `minimal-change-discipline`, `codebase-onboarding` |
+| **Technical Writer** (optional, project-specific) | `technical-writing` | `requirement-quality-checklist`, `spec-clarify` |
+| **CEO Orchestrator** | (meta — slash-spawned primitives) | `parallelization-by-default`, `help-me`, `cookbook-advisor` |
+| **Principal Incident Commander** (PLAN-074 Wave 1c — VETO floor) | `incident-management` | — |
+| **Principal Identity and Trust Architect** (PLAN-074 Wave 1c — VETO floor) | `identity-and-trust-architecture` | `security-and-auth` |
+| **Principal Threat Detection Engineer** (PLAN-074 Wave 1c — VETO floor on detection coverage / FPR drift) | `security-and-auth` | `chaos-and-resilience` |
+| **LLM FinOps Architect** (PLAN-074 Wave 1c — Advisory, NO VETO per ADR-052 amendment) | `llm-routing-and-finops` | — |
+
+### Domain skill map (optional — add entries for domain profiles you install)
+
+When you install a domain profile (e.g. `--profile core,fintech`), add its skills and the archetypes that own them to this section. For the fintech example, see `.claude/skills/domains/fintech/team-personas.md`.
+
+#### Domain `academic-humanities` (research workflows in the humanities)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| IRB/Ethics Coordinator | `psychologist` | Human-subjects research ethics |
+| Editor | `narratologist` | Citation policy, accuracy standards |
+| Reference Librarian | `anthropologist` | Archive access, source licensing |
+| Research Historian | `historian` | Primary source analysis, archival access |
+| Cultural Geographer | `geographer` | Spatial analysis, regional studies |
+
+#### Domain `business-support` (analytics, executive reporting, finance ops)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Support Operations Lead | `support-responder` | Customer support triage |
+| Analytics Engineer | `analytics-reporter` | Support + operations analytics |
+| Executive Reporting Lead | `executive-summary` | Board/exec-level reporting |
+| Finance Operations Specialist | `finance-tracker` | Budget tracking, expense management |
+
+#### Domain `civil-engineering` (structural, geotechnical, hydraulic, transport)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Civil Engineer | `civil-engineer` | IBC/ASCE-7/NBR-ABNT multi-jurisdiction structural |
+
+#### Domain `devrel` (developer relations)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Developer Advocate | `developer-advocate` | SDK docs, developer experience, community |
+
+#### Domain `edtech` (educational technology)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Student Privacy Engineer | `student-data-privacy` | FERPA/COPPA compliance |
+| Assessment Integrity Engineer | `assessment-integrity` | Test validity, anti-cheating |
+| Learning Analytics Engineer | `learning-analytics` | LMS analytics, outcome measurement |
+| Study Abroad Advisor | `study-abroad-advisory` | International application, visa, PII |
+
+#### Domain `embedded` (firmware and MCU development)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Embedded Firmware Engineer | `embedded-firmware` | FreeRTOS/Zephyr, OTA, secure boot |
+
+#### Domain `finance-accounting` (financial reporting, FP&A, tax)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Bookkeeper/Controller | `bookkeeper-controller` | GL, close, reconciliation |
+| Financial Analyst | `financial-analyst` | Variance, reporting |
+| FP&A Analyst | `fpa-analyst` | Budgeting, forecasting |
+| Tax Strategist | `tax-strategist` | Corporate tax compliance |
+
+#### Domain `fintech` (crypto/trading platform — reference team)
+
+Full team persona map at `.claude/skills/domains/fintech/team-personas.md`.
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Staff Quant | `financial-correctness-and-math` | PnL, float discipline (VETO) |
+| Exchange Integration Architect | `exchange-api-integration` | 21-adapter WS/REST gateway |
+| Real-Time Systems Engineer | `real-time-market-systems` | WebSocket, IPC, order-book |
+| Trading Engineer | `trading-execution` | SOR, market-making, execution quality |
+| Prediction Markets Engineer | `prediction-markets` | Resolution logic, probability math |
+| Frontend Data Engineer | `financial-display` | Price formatting, decimal precision |
+| Exchange Onboarding Lead | `exchange-onboarding-playbook` | New exchange runbook |
+| Blockchain Security Auditor | `blockchain-security-audit` | EVM/Solidity EthTrust/OWASP SCSVS |
+| Equity Research Analyst | `equity-research` | Fundamental + quant equity, DCF, factors |
+| Smart Contract Engineer | `solidity-smart-contracts` | Solidity authoring, reentrancy, DeFi |
+
+#### Domain `government` (public-sector digital transformation)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Government A11y Engineer | `accessibility-section-508` | Section 508, FedRAMP accessibility |
+| FOIA Compliance Officer | `foia-and-records` | Records management, public records law |
+| Procurement Integrity Officer | `public-procurement` | Nova Lei de Licitações, FedRAMP procurement |
+| Public-Sector Presales Engineer | `digital-presales` | Government bid lifecycle, FedRAMP/GovRAMP/FISMA |
+
+#### Domain `healthcare` (healthcare customer service + marketing compliance)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Healthcare CS Specialist | `healthcare-customer-service` | HIPAA-aware support |
+| Marketing Compliance Reviewer | `marketing-compliance` | FDA/FTC healthcare marketing rules |
+
+#### Domain `hospitality` (hotel/resort guest experience)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Guest Services Engineer | `guest-services` | PII-touching PMS integrations, guest journey |
+
+#### Domain `hr` (human resources)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| HR Onboarding Specialist | `hr-onboarding` | New-hire flows, compliance checklists |
+| Recruitment Specialist | `recruitment-specialist` | ATS, sourcing, bias-reduction |
+
+#### Domain `i18n-business` (international business + localization)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Cultural Intelligence Lead | `cultural-intelligence` | Cross-cultural business communication |
+| French Consulting Lead | `french-consulting` | French-market business practices |
+| Korean Business Lead | `korean-business` | Korean-market protocol, chaebols |
+| Language Translator | `language-translator` | MT quality, terminology management |
+
+#### Domain `identity-systems` (identity graph, customer 360)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Identity Graph Operator | `identity-graph-operator` | Entity resolution, PII graph |
+
+#### Domain `legal` (contract + matter management)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Client Intake Specialist | `client-intake` | Matter intake, conflict check |
+| Document Review Analyst | `document-review` | eDiscovery, privilege review |
+| Legal Billing Engineer | `legal-billing` | LEDES, matter billing, AFA |
+
+#### Domain `lgpd-heavy-saas` (LGPD-intensive SaaS platforms)
+
+Persona map at `.claude/skills/domains/lgpd-heavy-saas/team-personas.md`.
+
+#### Domain `marketing-global` (content, social, SEO, growth)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| SEO Specialist | `seo-specialist` | Technical + content SEO |
+| Content Creator | `content-creator` | Blog, long-form, editorial |
+| Social Media Strategist | `social-media-strategist` | Cross-channel social strategy |
+| Instagram Curator | `instagram-curator` | IG content strategy, Reels |
+| LinkedIn Content Creator | `linkedin-content-creator` | B2B LinkedIn, thought leadership |
+| TikTok Strategist | `tiktok-strategist` | Short-form video, creator collaboration |
+| Twitter/X Engager | `twitter-engager` | Real-time engagement, community |
+| Growth Hacker | `growth-hacker` | PLG, viral loops, activation |
+| Podcast Strategist | `podcast-strategist` | Show format, distribution, sponsorship |
+| Reddit Community Builder | `reddit-community-builder` | Subreddit growth, community trust |
+| Book Co-Author | `book-co-author` | Long-form narrative, editorial pipeline |
+| Carousel Growth Engine | `carousel-growth-engine` | Carousel post design, engagement |
+| App Store Optimizer | `app-store-optimizer` | ASO, keyword ranking, conversion |
+| Video Optimization Specialist | `video-optimization-specialist` | YouTube SEO, retention, CTR |
+| AI Citation Strategist | `ai-citation-strategist` | LLM citation positioning |
+| Agentic Search Optimizer | `agentic-search-optimizer` | AI search / AEO strategy |
+
+#### Domain `mobile` (mobile app development)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Mobile App Builder | `mobile-app-builder` | React Native / Flutter cross-platform |
+
+#### Domain `paid-media` (paid advertising operations)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Paid Media Auditor | `auditor` | Account audit, waste detection |
+| Creative Strategist | `creative-strategist` | Ad creative testing, messaging |
+| Paid Social Strategist | `paid-social-strategist` | Meta/TikTok campaign management |
+| PPC Strategist | `ppc-strategist` | Google/Bing search ads |
+| Programmatic Buyer | `programmatic-buyer` | DSP, audience segmentation |
+| Search Query Analyst | `search-query-analyst` | SQR analysis, negative lists |
+| Tracking Specialist | `tracking-specialist` | GTM, pixel, attribution |
+
+#### Domain `project-management` (delivery, studio ops)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Experiment Tracker | `experiment-tracker` | A/B test governance, KPI tracking |
+| Project Shepherd | `project-shepherd` | Delivery management, risk tracking |
+| Studio Operations Lead | `studio-operations` | Creative studio workflow |
+| Studio Producer | `studio-producer` | Campaign production management |
+
+#### Domain `real-estate-finance` (real estate transactions)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Buyer/Seller Agent Advisor | `buyer-seller-agent` | Transaction lifecycle, disclosure |
+| Loan Officer Assistant | `loan-officer-assistant` | Mortgage workflow, underwriting |
+
+#### Domain `retail` (e-commerce and physical retail)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Customer Returns Specialist | `customer-returns` | RMA workflow, policy enforcement, PII |
+
+#### Domain `saas-platforms` (SaaS platform engineering — Salesforce, Filament, CMS)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Platform Architect | `salesforce-architect` | Salesforce CRM, Apex, LWC |
+| CMS Developer | `cms-developer` | Headless CMS, content API |
+| Filament Specialist | `filament-specialist` | Laravel/Filament admin panels |
+
+#### Domain `sales` (B2B SaaS sales operations)
+
+Full team persona map at `.claude/skills/domains/sales/team-personas.md`.
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Revenue Operations Analyst | `outbound-strategist` | Outbound prospecting, sequence design |
+| Account Executive Lead | `discovery-coach` | Discovery facilitation, qualification |
+| Account Executive Lead | `proposal-strategist` | Deal proposals, competitive positioning |
+| Sales Coach | `sales-coach` | Rep coaching, call review, ramp |
+| Sales Development Rep | `sales-outreach` | Outbound cadence, channel mix |
+| Account Executive | `deal-strategist` | Deal structure, negotiation |
+| Account Executive | `account-strategist` | Strategic account planning |
+| Sales Engineer | `sales-engineer` | Technical demo, POC scoping |
+| Pipeline Analyst | `pipeline-analyst` | Funnel analytics, stage conversion |
+
+#### Domain `supply-chain` (supply chain management)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Supply Chain Strategist | `supply-chain-strategist` | Network optimization, demand planning |
+
+#### Domain `trading-hft` (high-frequency trading infrastructure)
+
+Full team persona map at `.claude/skills/domains/trading-hft/team-personas.md`.
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| HFT Kill-Switch Engineer | `kill-switches` | Circuit breakers, emergency halt |
+| HFT Latency Engineer | `latency-budgets` | Microsecond budget, profiling |
+| HFT Order Router | `order-routing` | Smart order routing, venue selection |
+
+#### Domain `training-l-and-d` (corporate learning and development)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Corporate Training Designer | `corporate-training-designer` | ILT/VILT, LMS, instructional design |
+
+#### Domain `voice-ai` (voice AI and conversational AI)
+
+| Archetype | Primary skill | Notes |
+|-----------|---------------|-------|
+| Voice AI Integration Engineer | `voice-ai-integration` | STT/TTS pipelines, telephony, wake-word |
+
+#### Domain `community` (curated imports, opt-in — PLAN-033 / ADR-060)
+
+The `community` domain holds externally-curated skills imported from
+upstream corpora (e.g. `sickn33/antigravity-awesome-skills`,
+MIT / CC-BY-4.0 / Apache-2.0) under the SP-NNN chain + rubric gate.
+Each imported skill carries `source:` + `license:` + `sp_chain:`
+frontmatter (see `.claude/skills/domains/community/NOTICE.md` for the
+attribution ledger). Archetypes:
+
+| Archetype | Primary skill | Secondary |
+|-----------|---------------|-----------|
+| **Community Researcher** | (meta) | (varies per imported skill) |
+| **Community Skill Steward** | (varies) | `architecture-decisions` |
+
+Kill-switch: `CEO_ANTIGRAVITY_SYNC=0` (operational convention; the
+scripts are Owner-invoked, so this signals "no imports right now" to
+CI / monitoring). Per-skill lifecycle in
+`.claude/skills/domains/community/team-personas.md`.
+
+---
+
+## ROUTING TABLE (MANDATORY — CEO MUST follow)
+
+> **Rule:** IF the work falls into a category below → SPAWN the listed agent(s).
+> **The CEO NEVER does the specialist's work.** The CEO orchestrates, the specialist executes.
+
+> **Dispatch path** (PLAN-061 / ADR-082): the dispatcher
+> (`.claude/scripts/inject-agent-context.sh`) resolves rail per
+> archetype default. `Staff Code Reviewer` runs on the **native** rail
+> (full tool grant works empirically; preserves ADR-052 VETO floor).
+> All other archetypes default to **mitigated** (Task dispatched via
+> `subagent_type=general-purpose` with persona injected — bypasses the
+> H4 rail anomaly per ADR-080). Operator overrides:
+> `--dispatch=native|mitigated`, `CEO_DISPATCHER_MODE=native|mitigated`,
+> kill-switch `CEO_MITIGATION_DISABLE=1` (force native universally).
+
+> **Spawn-depth doctrine (PLAN-135 D6+H11 — flat hierarchy):** the
+> harness now lets subagents nest up to **5 levels deep**. This
+> framework's doctrine is a **FLAT spawn topology**: the CEO spawns
+> specialists (depth 1); specialists do **NOT** spawn subagents of
+> their own. Every row in this table is a CEO→specialist dispatch —
+> the governance rail (`check_agent_spawn.py` spawn contract +
+> `audit_log.py` observer) is designed and tested around flat spawns;
+> a 5-deep ungoverned fan-out multiplies quota burn geometrically and
+> executes below the per-spawn governance the rail provides at
+> depth 1. `CEO_MAX_SPAWN_DEPTH` is **RESERVED** as the future
+> enforcement knob (configurable depth ceiling read from
+> `agent_id`/`parent_agent_id` chains in the hook payloads).
+> **Enforcement is harvest item H11 — DEFERRED**: until it lands, the
+> depth ceiling is doctrine (this paragraph), not a hook. If a task
+> genuinely needs a sub-decomposition, the specialist returns
+> `BLOCKED`/`NEEDS_CONTEXT` and the CEO re-plans the split at depth 1
+> (see §AGENT SPAWN PROTOCOL Step 5 escalation rules — "too large →
+> split into file-disjoint units").
+
+> **Agent Teams pre-positioning (PLAN-135 D6 — research-preview, NOT
+> adopted):** Agent Teams (peer-teammate topology; `TeammateIdle` /
+> `TaskCompleted` lifecycle events) is a **research-preview** surface.
+> Status: **NOT adopted** by this framework; never default-on while in
+> research-preview. Topology note: a peer team is NOT a spawn tree —
+> teammates are siblings coordinating laterally, which bypasses the
+> CEO→specialist routing this table encodes, so adoption is a
+> governance decision (own ADR + prereg pilot), not a convenience
+> toggle. Candidate prereg pilot shape (recorded for a future plan;
+> requires its own pre-registered falsifier per the PLAN-134
+> discipline): map 2-3 archetypes (e.g. Staff Code Reviewer + Security
+> Engineer) onto teammates, treat **`TaskCompleted` exit-2 ≈ native
+> VETO** (a non-zero teammate exit gates the merge exactly like a
+> `BLOCK` verdict), and route `TeammateIdle`/`TaskCompleted` through
+> the spawn-governance + audit rail so the experiment is observable
+> before any adoption decision.
+
+| Work type | Agent archetype | Skill to load | Approver | Dispatch path |
+|-----------|-----------------|---------------|----------|---------------|
+| API design, contracts, OpenAPI | **Staff Backend Engineer** + **Staff Code Reviewer** | `public-api-design` + `code-review-checklist` | Code Reviewer | mitigated + native |
+| Security, auth, encryption, threat modeling | **Security Engineer** | `security-and-auth` | VP Operations | mitigated |
+| Performance, event loop, memory, GC | **Principal Performance Engineer** | `performance-engineering` | VP Engineering | mitigated |
+| Database schema, migrations, RLS | **Principal Data Engineer** | `data-schema-design` | VP Engineering | mitigated |
+| Resilience, circuit breakers, failure modes | **Chaos Engineer** | `chaos-and-resilience` | VP Operations | mitigated |
+| Tests, QA, edge cases, regression | **Principal QA Architect** | `testing-strategy` | Code Reviewer | mitigated |
+| Real-time systems, WebSocket, state machines | **Real-Time Systems Engineer** | `state-machines-and-invariants` | VP Engineering | mitigated |
+| Billing, payments, subscriptions | **Billing Engineer** | `monetization-and-billing` | VP Product | mitigated |
+| Compliance, privacy, LGPD/GDPR, PII dataflow, consent lifecycle, DPO/DSR, ANPD reporting | **Compliance Specialist** | `compliance-lgpd` + `core/pii-data-flow` + `core/consent-lifecycle` + `core/dpo-reporting` (Phase 0a / ADR-120) | Compliance Specialist | mitigated |
+| CI/CD, Docker, deploys, platform | **DevOps Engineer** | `devops-ci-cd` | VP Operations | mitigated |
+| Growth, onboarding, conversion | **Growth Engineer** | `growth-and-launch` | VP Product | mitigated |
+| Architecture (3+ modules touched) | **VP Engineering** | `architecture-decisions` | Owner | mitigated |
+| AI integration, LLM prompts, AI safety | **Security Engineer** + **Staff Backend Engineer** | `ai-llm-orchestration` + `security-and-auth` | Security Engineer | mitigated |
+| Code review (EVERY change) | **Staff Code Reviewer** | `code-review-checklist` | Code Reviewer | native |
+| Frontend work | see `.claude/frontend-team.md` | — | Frontend leads | mitigated |
+| Curated skill import (new community SKILL.md) | **Community Skill Steward** | — (uses `.claude/scripts/import-skill.py`) | Owner (per-skill SP-NNN sign) | mitigated |
+| Mechanism selection ("which artifact for X?") | **VP Engineering** + **Staff Code Reviewer** | `architecture-decisions` + `code-review-checklist` | VP Engineering | mitigated + native |
+| Autonomous-loop parallelism (opt-in) | **CEO + Swarm Coordinator** | (uses .claude/scripts/swarm/coordinator.py) | VP Engineering | mitigated |
+| Live incident triage / severity classification / paging policy / on-call / runbook authoring | **Principal Incident Commander** | `incident-management` | Incident Commander (VETO on severity / all-clear / detection-pager surface during incident) | native |
+| Identity / token lifecycle (JWT / OAuth / OIDC) / RBAC-ABAC / mTLS S2S / IdP integration / RLS / session lifecycle | **Principal Identity and Trust Architect** | `identity-and-trust-architecture` | Identity & Trust Architect (VETO on credential issue/validate/rotate/revoke; cross-service trust brokerage; role/scope assignment) | native |
+| Detection-as-code / SIEM rules / MITRE ATT&CK coverage / FPR audit / alert dedup / threat-hunt playbooks / purple-team / log-source coverage | **Principal Threat Detection Engineer** | `security-and-auth` (§Detection-as-Code corpus) | Threat Detection Engineer (VETO on detection-coverage gaps + noisy-rule deployments) | native |
+| LLM cost envelope / model routing / per-plan token budgets / burn-rate / parent-inheritance trap / tier policy review / sub-agent dispatch model selection | **LLM FinOps Architect** | `llm-routing-and-finops` | VP Operations (advisory — NO VETO per ADR-052/ADR-064 amendment) | mitigated |
+
+| Documentation authoring, ADR writing, README, changelog, runbook | **Technical Writer** | `technical-writing` | VP Engineering | mitigated |
+| Requirements quality review, plan spec validation | **Principal QA Architect** | `requirement-quality-checklist` | Code Reviewer | mitigated |
+| Plan ambiguity reduction, spec clarification (manual /spawn) | **Principal QA Architect** + **VP Product** | `spec-clarify` | VP Product | mitigated |
+| Pair-rail cross-LLM review (L2+ tasks, security-critical) | **Staff Code Reviewer** | `cross-llm-pair-review` | Code Reviewer | native |
+| MCP server authoring, tool registration, schema design | **Staff Backend Engineer** | `mcp-server-authoring` | VP Engineering | mitigated |
+| Evidence-based QA sign-off, test signal interpretation | **Principal QA Architect** | `evidence-based-qa` | Code Reviewer | mitigated |
+| LSP-anchored code analysis, type-safety, refactor analysis | **VP Engineering** | `code-intelligence-lsp` | VP Engineering | mitigated |
+| Git workflow, branch strategy, commit message, PR / release tag | **DevOps Engineer** | `git-workflow-discipline` | VP Engineering | mitigated |
+| Codebase orientation before modification (new domain/module) | **Refactoring Lead** | `codebase-onboarding` | VP Engineering | mitigated |
+| Change-scope discipline (minimal-change enforcement) | **Staff Code Reviewer** | `minimal-change-discipline` | Code Reviewer | native |
+| Cross-artifact consistency audit (coverage-audit /spawn) | **Principal QA Architect** | `coverage-audit` | Code Reviewer | mitigated |
+| Task decomposition + parallel sub-agent dispatch | **CEO Orchestrator** | `parallelization-by-default` | VP Engineering | — |
+| Owner contextual help ("which skill/command for X?") | **CEO Orchestrator** | `help-me` | — | — |
+| Anthropic Cookbook pattern advisory (COOK-P1..P4) | **CEO Orchestrator** | `cookbook-advisor` | — | — |
+
+Extend this table with domain-specific routes when you install a domain profile.
+
+---
+
+## CEO orchestrator model tier (PLAN-048 Phase 1)
+
+> **Status:** candidate routing rule — see `docs/CEO-MODEL-ROUTING.md`.
+> `CEO_MODEL_DOWNSHIFT=0` reverts to Opus-always (experiment kill-switch).
+
+**Default CEO model:** Sonnet 4.6. **Upgrade to Opus 4.8 upfront** if ANY:
+
+| # | Condition | Why |
+|---|---|---|
+| a | Plan frontmatter `level: L3` or higher | L3+ blast radius requires deep protocol compliance |
+| b | Session tag ∈ `{L3+-plan-execution, debate-round, brainstorm, ceremony}` | Empirically spawn-heavy class (PLAN-048 baseline N=8) |
+| c | Canonical-edit path declared in session scope | Governance-critical paths need protocol rigor |
+| d | VETO-protected domain touched | auth / financial-math / token handling |
+| e | Expected `spawn_count` ≥ 3 by session-tag heuristic | Multi-phase plan-execution pattern |
+
+**Invariants NOT affected:**
+- **VETO roles (code-reviewer, security-engineer) ALWAYS Opus 4.8.**
+  ADR-052 VETO floor is hardcoded in dispatcher. This rule only
+  shifts the CEO orchestrator identity.
+- **Sub-agent tier unchanged.** ADR-052 role-to-model mapping is
+  independent of CEO tier.
+- **Kill-switch `CEO_MODEL_DOWNSHIFT=0` preserves adopter escape.**
+
+---
+
+## AGENT SPAWN PROTOCOL (MANDATORY — read every session)
+
+> **The old template approach was cosmetic** — calling a generic LLM by a persona name did nothing unless that persona's actual skill content was in the prompt. This protocol is what makes agents real.
+
+### Step 0: FILE ASSIGNMENT (BEFORE spawning — anti-collision)
+
+> **ABSOLUTE RULE:** Two agents NEVER edit the same file in parallel. Violation = lost work. No exceptions.
+
+Before spawning 2+ agents in parallel, the CEO MUST:
+
+1. **List the files** each agent will touch
+2. **Verify zero overlap** — if two agents need the same file → run SEQUENTIAL
+3. **Declare the file assignment** in each agent's prompt:
+   ```
+   YOUR FILES (ONLY YOU can edit these):
+   - src/path/to/file1.ts
+   - src/path/to/file2.ts
+
+   FORBIDDEN FILES (another agent is editing):
+   - src/path/to/other.ts (AgentX is editing)
+   - src/path/to/shared.ts (AgentY is editing)
+   ```
+4. If an agent NEEDS to read (not edit) a file another agent is editing → OK, read is safe
+5. If during execution the agent discovers it needs to edit a forbidden file → STOP and report to CEO
+
+### Parallelism modes
+
+| Mode | When to use | Collision risk |
+|------|------------|----------------|
+| **No worktree** (default) | Agents edit DIFFERENT files | ZERO if file assignment correct |
+| **With worktree** (`isolation: "worktree"`) | Agents may touch the same files | LOW, manual merge after |
+| **Sequential** | Agents MUST edit the same file | ZERO (one waits for the other) |
+
+### Decision rule
+
+- **0 files in common** → parallel WITHOUT worktree (fastest)
+- **1-3 files in common** → SEQUENTIAL (safest)
+- **4+ files in common** → probably 1 task, not 2 — collapse them
+
+### Step 1: Read the agent profile
+
+The CEO reads the agent block in this file (team.md) to obtain: name, title, background, focus, superpower, quirks, red flags, expected output, mantra.
+
+### Step 2: Read the agent's skill
+
+The CEO reads `.claude/skills/<tier>/<skill-name>/SKILL.md` (path determined by the SKILL MAP above). Where `<tier>` is `core`, `frontend`, or `domains/<domain>/skills`.
+
+### Step 3: Build the prompt with BOTH
+
+```
+Agent tool → prompt containing:
+
+1. PERSONA (copied from team.md)
+2. SKILL CONTENT (copied from SKILL.md, OR SKILL REFERENCE per ADR-051)
+3. SPEC CONTEXT (optional, if brainstorm ran — ADR-058)
+4. FILE ASSIGNMENT (Step 0 — files they may/may not edit)
+5. TASK with acceptance criterion
+6. CONSTRAINTS (what NOT to do)
+7. EXPECTED OUTPUT FORMAT
+```
+
+### Template prompt
+
+```
+PERSONA: {Name} — {Title} of {{PROJECT_NAME}}
+BACKGROUND: {Full background}
+FOCUS: {Focus areas}
+RED FLAGS: {What to detect}
+MANTRA: {Mantra}
+
+## SKILL CONTENT
+SKILL LOADED: {skill name}
+{Full SKILL.md content — rules, checklists, patterns}
+
+## SPEC CONTEXT   (optional — present only if pre-plan-brainstorm ran, ADR-058)
+SPEC FROM: .claude/plans/PLAN-NNN/spec.md sha256={hex}
+{Full spec.md content — stakeholders, success criteria, anti-goals,
+ constraints, assumptions, known unknowns, tradeoffs, outcomes,
+ open questions. For debate Round 1, this context replaces the
+ "what does the Owner want" conversation.}
+
+## FILE ASSIGNMENT
+- MAY edit: {list of files}
+- MAY NOT edit: {list of files another agent is editing}
+- If you need to edit a forbidden file: STOP and report.
+
+## TASK
+{Clear task description}
+ACCEPTANCE CRITERION: {How to know you're done}
+
+## CONSTRAINTS
+{What not to do}
+
+## OUTPUT FORMAT
+{Expected structure}
+```
+
+**Notes on `## SPEC CONTEXT`:**
+
+- Present only when `spec_ref:` is set in the plan's frontmatter
+  (PLAN-SCHEMA §Optional frontmatter, ADR-058).
+- Mirrors the `## SKILL CONTENT` / `## SKILL REFERENCE` pattern:
+  either inline (content embedded) OR by-reference (sha256 hash)
+  per ADR-051. Inline is simpler; reference is cache-friendlier
+  for large specs.
+- For debate Round 1 agents, the spec replaces the implicit
+  "what does Owner want?" question with an explicit contract
+  the agent can cite in its critique.
+- Absent when `CEO_BRAINSTORM_GATE=0` is set OR plan is L1-L2
+  OR plan has no `spec_ref:`.
+
+### Step 4: Validate the output
+
+When the agent returns, the CEO verifies:
+
+- [ ] The agent edited ONLY files from its file assignment?
+- [ ] The output reflects knowledge of the skill? (uses terms/patterns from the skill)
+- [ ] The output follows the requested format?
+- [ ] The output is verifiable against the code? (not hallucinated)
+
+If NO → Strike for the agent + CEO retries with more context.
+
+### Step 5: Worker return-status contract (PLAN-115 WS-C)
+
+The §Step 4 checklist governs what the CEO verifies; this step governs what the
+**worker must report**. Every spawned worker returns one of four statuses (this
+is the MAP-side companion to the REDUCE-side verification doctrine in
+`docs/triage-reduce-protocol.md`, ADR-141):
+
+| Status | Meaning | Controller (CEO) action |
+|--------|---------|--------------------------|
+| `DONE` | Task complete, acceptance criteria met, verified against code | Run §Step 4 validation; accept |
+| `DONE_WITH_CONCERNS` | Complete, but the worker flags a risk/assumption it could not resolve | Validate + surface the concern; do not silently drop it |
+| `NEEDS_CONTEXT` | Blocked on missing information, not on capability | Re-dispatch the **same model** with the missing context added (not a more capable one) |
+| `BLOCKED` | Cannot proceed (capability, scope, or a wrong plan) | Diagnose the cause, then escalate per the table below |
+
+**Escalation rules (never force a same-model retry without a change):**
+
+- **Context problem** (worker lacked a file / spec / prior decision) →
+  re-dispatch the **same model** with the missing context. Same model + same
+  prompt + same gap = same failure.
+- **Needs more reasoning** (genuine difficulty, not missing context) →
+  re-dispatch on a **more capable model** (per ADR-052 tier policy).
+- **Too large** (the unit is bigger than one worker should own) → **split** it
+  into smaller file-disjoint units (see §Step 0 file assignment).
+- **Plan is wrong** (the task as specified cannot be right) → **escalate to the
+  Owner**; do not have the worker improvise a different plan.
+
+### Step 6: Verify, don't trust the worker report (PLAN-115 WS-C)
+
+A worker's claim ("done, all tests pass") is a claim to **verify against the
+code**, not a fact to accept — the same discipline the `receiving-review` skill
+applies to feedback. The reviewer/controller re-inspects the actual files and
+runs the checks before accepting a `DONE`.
+
+The companion rule — *provide the worker the full task text rather than a
+pointer to "go read the plan"* — rests on the existing spawn contract:
+`check_agent_spawn.py` requires **skill material** (`## SKILL CONTENT` inline OR a
+valid `## SKILL REFERENCE`), and the AGENT-PROFILE template (PROTOCOL §Spawn
+Protocol Step 3) is where the full task text is carried. Full task-text
+*completeness* is protocol doctrine (Step 3), not a hook assertion — it is not
+re-implemented here; see §Step 3 and `check_agent_spawn.py`.
+
+---
+
+## GOVERNANCE RULES
+
+### Approvals and vetoes
+
+#### Code Reviewer VETO (any merge) — BLOCK if ANY:
+- [ ] Type checker has errors (stack-specific: `tsc --noEmit`, `mypy`, `go vet`, etc.)
+- [ ] Test suite has failures
+- [ ] New code without corresponding test
+- [ ] Inconsistent naming with existing patterns
+- [ ] Functions above the project's agreed line-count limit without decomposition justification
+- [ ] Missing error handling on async operations
+
+#### Staff Domain Expert VETO (per domain) — BLOCK if ANY:
+Domain-specific rules defined per project. See `.claude/skills/domains/<domain>/pitfalls.yaml` for the full list of domain-specific blockers. Examples:
+- Fintech: float arithmetic on financial values, missing boundary tests on math, missing invariant checks
+- Healthcare: PHI leakage in logs, missing audit trails, weak encryption
+- Auth-heavy SaaS: missing CSRF protection, JWT without proper validation, rate limiting gaps
+
+#### VP Engineering APPROVAL (architecture — 3+ modules touched):
+- [ ] ADR documented with trade-off analysis
+- [ ] Blast radius assessed (which modules affected)
+- [ ] Scales to N× current load without rewrite
+
+#### VP Operations APPROVAL (deploys):
+- [ ] Health check endpoints verified
+- [ ] Rollback plan documented
+- [ ] Smoke test defined (not just a single endpoint)
+
+### 3-Strike Policy
+
+Every named agent starts at 0/3 strikes. See `.claude/agent-metrics.md` for the tracking template.
+
+A strike is recorded when the agent produces:
+
+- A **factual error** that can be verified against the code (claims file X exists when it doesn't)
+- A **skill violation** (the security agent forgets auth; the financial agent uses floats)
+- **Incomplete output** (says "done" but key files are missing)
+- A **regression** (their fix breaks existing tests)
+
+NOT a strike:
+
+- A different-but-valid approach (if it works, the disagreement is taste)
+- An error caused by a bad prompt from the CEO (the CEO failed, not the agent)
+
+Consequences:
+
+- **1/3** — Warning in `.claude/agent-metrics.md`, "ATTENTION" flag in the next spawn prompt
+- **2/3** — Supervised mode: another agent reviews every output
+- **3/3** — Fired. Persona is rewritten. A new agent with a new name replaces them.
+
+---
+
+## Extending this team for your project
+
+1. **Add concrete personas.** The archetype tables above work, but vivid personas (with backgrounds, quirks, and mantras) produce more consistent outputs. See `.claude/skills/domains/fintech/team-personas.md` for a worked example with 18 backend personas.
+
+2. **Add domain skills.** When you install a domain profile (e.g. `--profile core,fintech`), add its skills and the archetypes that own them to the SKILL MAP and ROUTING TABLE sections above.
+
+3. **Add domain VETOes.** Every project has 1-3 critical domains that warrant a VETO holder. Examples: financial math, PHI/PII, auth, infrastructure. Add them to the "Staff Domain Expert" section.
+
+4. **Customize the ROUTING TABLE** to reflect the work types your project actually does. The CEO uses this table to route work — if a work type isn't in the table, the CEO has to improvise, which is worse.
+
+5. **Define stack-specific tooling** in the Code Reviewer VETO (`tsc` for TypeScript, `mypy` for Python, `go vet` for Go, etc.).