agent-forge

Accessibility specialist for Microsoft platform UIs. TRIGGER when: user mentions Power Apps canvas, Power Apps model-driven, Power Pages, Blazor, Office add-in UI, Teams tab UI, WCAG, a11y, accessible, screen reader, keyboard nav, ARIA, focus management, color contrast, Accessibility Insights, axe DevTools, or invokes /accessibility. Codifies opinions: WCAG 2.2 AA is the floor, native HTML over ARIA, keyboard-first then screen-reader testing, manual coverage non-negotiable. Reads from standards/references/quality/accessibility-wcag.md (canonical checklist). DO NOT TRIGGER for non-UI workloads (backend services, data pipelines, IaC), Power Platform governance broadly (use powerplatform-architect), or .NET platform patterns broadly (use dotnet-architect).

AI and copilot architecture specialist. TRIGGER when: user needs Microsoft Foundry (Azure AI Foundry), Azure OpenAI Service, Copilot Studio custom copilots, AI agent architecture, RAG patterns, prompt engineering, vector stores (AI Search, pgvector), model catalog, prompt flow, AI evaluation, tracing, responsible AI, content safety, function calling, multi-agent orchestration, Foundry IQ, or invokes /ai-architect. Designs AI solutions on Microsoft Foundry with guardrails, responsible AI principles, and enterprise-grade observability. Fetches latest documentation from Microsoft Learn MCP. Produces agent architectures, RAG pipeline designs, Foundry project structures, and prompt engineering frameworks. DO NOT TRIGGER for Power Platform without AI focus (use powerplatform-architect), data engineering only (use data-architect), or general Azure (use azure-architect).

Architecture workbook generator for Microsoft projects. TRIGGER when: user needs Architecture Decision Records (ADR), effort estimation workbooks, RAID logs, solution design documents, or test strategy workbooks, or invokes /artifacts. Generates structured Excel workbooks using the xlsx skill. DO NOT TRIGGER for HLD/LLD documents (use docs), tech specs (use spec), or design work (use specialist skills).

Azure PaaS and IaaS architecture specialist. TRIGGER when: user needs Azure architecture design, App Service, Functions, APIM, Azure SQL, Cosmos DB, networking, identity, Azure-native patterns, or invokes /azure-architect. Designs Azure solutions validated against the Azure Well-Architected Framework (5 pillars). Fetches latest documentation from Microsoft Learn MCP. Produces C4 diagrams, integration patterns, and security architecture. DO NOT TRIGGER for Power Platform only (use powerplatform-architect), D365 only (use d365-architect), or containers only (use container-architect).

Azure SQL platform tuning specialist. TRIGGER when: user mentions Azure SQL Database, Azure SQL Managed Instance, Azure SQL Edge, Hyperscale tier selection, Query Store, partitioning strategy, Always Encrypted, audit or ledger, failover groups, geo-replication, Always On AGs on SQL VM, read-only routing, SqlClient connection resilience, or invokes /azure-sql-architect. Codifies opinions: Hyperscale > Premium for >1TB workloads, read-only routing > app-side R/W splitting, partition by date or tenant (never random hash), Query Store on every prod DB, Always Encrypted (deterministic) for searchable PII. DO NOT TRIGGER for Fabric, Databricks, Power BI, Purview, lakehouse design (use data-architect), Cosmos DB or PostgreSQL Flexible Server selection (use azure-architect), or app-side EF Core query patterns (use dotnet-architect and standards/references/coding-stack/ef-core-checklist.md).

CI/CD architecture specialist. TRIGGER when: user needs GitHub Actions workflow design, Azure DevOps Pipelines (YAML), GitOps for AKS or Container Apps, ring deployments, blue-green deployment, dependabot configuration, Conventional Commits enforcement, Terraform plan-in-PR workflow, environment promotion gates, secret rotation in pipelines, or invokes /cicd-architect. Codifies opinions: GitHub Actions for OSS-friendly orgs; Azure DevOps for enterprise with strict gates. GitOps for AKS/Container Apps (Flux or ArgoCD). Ring deployments (canary -> 1% -> 10% -> 50% -> 100%) for production releases. Blue-green for stateful workloads where ring isn't viable. Dependabot grouped by package ecosystem; auto-merge minor/patch with passing tests. Conventional Commits enforced via commitlint in PR. Terraform plan in PR comment via tfcmt or atlantis; apply only post-merge. Reads from standards/references/tooling/repo-baseline.md. DO NOT TRIGGER for Azure service selection (use azure-architect), infrastructure provisioning

Container and Kubernetes architecture specialist. TRIGGER when: user needs AKS cluster design, Container Apps, DAPR, GitOps with Flux, service mesh, Helm charts, container security, microservices orchestration, or invokes /container-architect. Designs container-native solutions on Azure validated against the Azure Well-Architected Framework. Fetches latest documentation from Microsoft Learn MCP. Produces cluster topologies, DAPR component diagrams, and GitOps pipeline designs. DO NOT TRIGGER for simple App Service deployments (use azure-architect), Power Platform (use powerplatform-architect), or D365 (use d365-architect).

Dynamics 365 architecture specialist. TRIGGER when: user needs D365 CE (Sales, Service, Field Service, Project Ops, Customer Insights), D365 F&O (Finance, SCM, HR, Warehouse), Business Central, X++ development, dual-write, virtual entities, Success by Design, or invokes /d365-architect. Designs Dynamics 365 solutions using Success by Design methodology. Fetches latest documentation from Microsoft Learn MCP. Produces entity models, integration patterns, and ISV solution assessments. DO NOT TRIGGER for Power Platform only (use powerplatform-architect), Azure PaaS only (use azure-architect), or data analytics only (use data-architect).

Data platform and analytics architecture specialist. TRIGGER when: user needs Microsoft Fabric, Azure Databricks, Power BI, data governance (Purview), medallion architecture, data lakehouse, ETL/ELT pipelines, data migration, DuckDB, SQLite, or invokes /data-architect. Designs data platform solutions validated against the Fabric WAF and Azure WAF principles. Fetches latest documentation from Microsoft Learn MCP. Produces data models, pipeline designs, and governance frameworks. DO NOT TRIGGER for OLTP database selection only (use azure-architect), D365 data entities only (use d365-architect), or Dataverse only (use powerplatform-architect).

Defender for Cloud depth specialist and Microsoft Sentinel SOC architect. TRIGGER when: user asks about Defender for Cloud plan selection (Server, App Service, SQL, Storage, Container, Key Vault, Resource Manager, DNS), secure score, Microsoft Cloud Security Benchmark compliance dashboard, Sentinel workspace design, Sentinel data connectors, analytics rules (scheduled, NRT, Fusion, ML, Microsoft security), hunting queries (KQL), watchlists, automation rules, SOAR playbooks via Logic Apps, threat intelligence integration (TAXII, MISP, custom feeds), MITRE ATT&CK mapping, SOC engineering, SIEM architecture, or invokes /defender-sentinel. DO NOT TRIGGER for general security posture, Key Vault patterns, private endpoints, network segmentation, or supply chain security (use security-architect for those). DO NOT TRIGGER for KQL fundamentals or Log Analytics workspace basics (use observability-architect). DO NOT TRIGGER for STRIDE-A threat modeling (use threat-model).

Engagement discovery and intake for Microsoft architecture projects. TRIGGER when: user starts a new architecture engagement, describes a project that needs Microsoft architecture, or invokes /discover. Gathers project context, business domain, organization size, existing Microsoft landscape, compliance requirements, timeline, and engagement type. Outputs a structured discovery brief that feeds into stack-select. DO NOT TRIGGER for stack selection (use stack-select) or design work (use specialist skills).

HLD and LLD document generator for Microsoft architecture projects. TRIGGER when: user needs High-Level Design documents, Low-Level Design documents, tech-design-first specifications (design.md), architecture documentation, or invokes /docs. Produces comprehensive design documentation in docx, pptx, pdf, or markdown format. DO NOT TRIGGER for workbook artifacts (use artifacts), tech specs per story (use spec), or design work (use specialist skills).

.NET architecture specialist. TRIGGER when: user needs .NET / C# implementation guidance, ASP.NET Core (Minimal APIs / Web APIs / Blazor), EF Core data access, Aspire distributed orchestration, .NET testing strategy, or invokes /dotnet-architect. Codifies preferred-stack opinions: .NET 8/9 LTS only, Result<T> over exceptions, Managed Identity over connection strings, TDD by default. Reads from standards/references/coding-stack/csharp-standards.md and ef-core-checklist.md. DO NOT TRIGGER for Azure infrastructure (use azure-architect), identity provisioning (use identity-architect), or .NET Framework legacy migration (use dotnet-modernization in Phase 4).

.NET Framework to .NET 8/9 LTS migration specialist. TRIGGER when: user mentions legacy .NET Framework 4.x, ASP.NET (non-Core), WCF services, EF6, ASP.NET Web Forms, "modernize", "upgrade", or ".NET Framework"; or when discover outputs an engagement classified as Modernization with a .NET Framework stack. Codifies migration strategy: strangler fig over big bang for any non-trivial codebase (greater than 50K LOC), .NET Upgrade Assistant (now superseded by GitHub Copilot modernization agent in VS 2026) for assessment phase, WCF to gRPC (internal) or HTTP+OpenAPI (external), EF6 to EF Core with explicit review of LINQ queries and lazy loading, Web Forms to Blazor Server (preferred), Windows Container as last resort when full migration is not viable for 12-18 months. DO NOT TRIGGER for greenfield .NET 8/9 work (use dotnet-architect), Azure infrastructure sizing (use azure-architect), or containerization strategy for already-modern apps (use container-architect).

Azure FinOps and cloud cost architecture specialist. TRIGGER when: user needs Azure cost analysis, pricing API integration, resource group spend breakdown, Reserved Instance or Savings Plan selection, tagging strategy design, FinOps Foundation framework implementation (Inform / Optimize / Operate), cost anomaly detection, showback or chargeback model design, right-sizing recommendations, budget alert configuration, or invokes /finops-architect. Always-on horizontal; runs after every vertical architect engagement alongside security-architect, identity-architect, and observability-architect. DO NOT TRIGGER for general Azure infrastructure design (use azure-architect), security posture review (use security-architect), IaC authoring (use iac-architect), or SLO / error-budget design (use sre-architect in Phase 4).

Infrastructure-as-Code architecture specialist. TRIGGER when: user needs IaC design or review, mentions Terraform, Bicep, AVM modules, state management, drift detection, plan/apply workflow, brownfield import, IaC policy / governance, or invokes /iac-architect. Codifies opinions: Terraform > Bicep for greenfield, AVM modules first, remote state with locking + RBAC, PR-gated apply, drift detection in CI nightly. Reads from standards/references/tooling/repo-baseline.md. DO NOT TRIGGER for Azure service selection (use azure-architect) or identity provisioning (use identity-architect).

Identity & access architecture specialist. TRIGGER when: user mentions Entra ID, Azure AD, B2C, External ID, Managed Identity, Workload Identity Federation, RBAC role design, Conditional Access, PIM, MFA strategy, OAuth flows for Microsoft identity, or invokes /identity-architect. Codifies opinions: Managed Identity > service principal, Entra ID > B2C unless CIAM, Workload Identity Federation > long-lived secrets, RBAC > access policies. Reads from standards/references/security/identity-decision-tree.md. DO NOT TRIGGER for Azure service selection (use azure-architect), security controls broadly (use security-architect), or Defender/Sentinel (use defender-sentinel in Phase 4).

Microsoft 365 platform extensibility specialist. TRIGGER when: user needs SharePoint Framework (SPFx) web parts or extensions, Teams apps (tabs, bots, message extensions, meeting apps), SharePoint customizations, Office add-ins (Word, Excel, Outlook, PowerPoint via Office.js), Viva extensions (Connections, Engage, Topics adapters), Graph Connectors for 3rd-party content in M365 search, Adaptive Cards in Outlook actionable messages or Teams, or invokes /m365-platform. Designs M365 surface extensions validated against SPFx + Teams Toolkit (Microsoft 365 Agents Toolkit) patterns. Fetches latest documentation from Microsoft Learn MCP. Stack-pinned: chains after powerplatform-architect when SPFx, Teams apps, or Office add-ins are in scope. DO NOT TRIGGER for Power Platform only (use powerplatform-architect), pure Graph API integration without M365 surface (use microsoft-graph), or Azure-only solutions (use azure-architect).

.NET MAUI cross-platform mobile and desktop architecture specialist. TRIGGER when: user needs cross-platform mobile design (iOS, Android, Windows, macOS), Shell navigation, MVVM with CommunityToolkit.Mvvm, Blazor Hybrid for web UI reuse, offline-first SQLite patterns, push notifications via Azure Notification Hubs, native platform interop, app store deployment via dotnet publish, or invokes /maui-architect. Codifies preferred-stack opinions: .NET MAUI over Xamarin.Forms, Shell over NavigationPage, CommunityToolkit.Mvvm over roll-your-own ObservableObject, offline-first by default, Key Vault Code Signing. Reads from standards/references/coding-stack/csharp-standards.md. DO NOT TRIGGER for web-only Blazor (use dotnet-architect), Azure infrastructure SKUs (use azure-architect), identity provisioning (use identity-architect), or .NET Framework legacy migration (use dotnet-modernization).

Microsoft Graph API integration specialist. TRIGGER when: user mentions Microsoft Graph, Graph API, Graph SDK, change notifications, webhooks against M365, $batch, delta query, throttling against Graph, beta vs v1.0 endpoints, Graph permission scopes (User.Read.All, Mail.Read, Sites.Selected), or any vertical needing Microsoft 365 data (mail, calendar, Teams, SharePoint, OneDrive, users, groups). Codifies opinions: application vs delegated permissions are not interchangeable, beta endpoints are prohibited in production without an ADR, batch and webhooks beat polling, throttling-aware retry with jitter is mandatory, $select is mandatory for PII minimization. Reads from standards/references/research/learn-mcp-patterns.md and standards/references/security/identity-decision-tree.md. DO NOT TRIGGER for Azure service selection (use azure-architect), identity primitive selection broadly (use identity-architect), Power Platform Graph connectors (use powerplatform-architect), or SharePoint Framework / Teams app develo

Observability architecture specialist. TRIGGER when: user needs App Insights instrumentation, Log Analytics workspace design, KQL query patterns, OpenTelemetry SDK setup, distributed tracing, structured logging, SLO design, dashboards (Azure Workbooks or Azure Managed Grafana), Azure Monitor alert rules, action groups, burn-rate alerting, or invokes /observability-architect. Codifies opinions: App Insights + Log Analytics + Azure Monitor on every component (non-negotiable), OpenTelemetry as wire format (AppInsights SDK only for legacy), SLO-driven alerting on burn rate not thresholds, distributed tracing required for any system with more than 2 services, structured JSON logs only. Reads from standards/references/coding-stack/preferred-stack.md and standards/references/quality/review-checklist.md. DO NOT TRIGGER for SRE error-budget governance (use sre-architect in Phase 4), Sentinel SOC analytics rules (use defender-sentinel in Phase 4), or FinOps cost-of-logs analysis (use finops-architect).

Microsoft Odin: a problem-solving agent for enterprise architecture. TRIGGER when: user shares a functional spec, technical spec, PRD, architecture problem, or design challenge. Also triggers for architecture review, deep design analysis, "how should I build X", design plan, implementation strategy, "odin", "design review", "ultrathink", "scalable design", "composable design", or when coding agents need architecture guidance. Odin receives a problem (spec, requirement, or challenge), uses AskUserQuestion to gather missing context, and arrives at a validated, scalable design using plan mode. Applies functional programming paradigm and DDD principles. Produces secure-by-design architectures that pass strong security reviews. DO NOT TRIGGER for document generation (use docs) or spec file creation (use spec).

Power Platform architecture specialist. TRIGGER when: user needs Power Apps (Canvas, Model-driven), Power Automate, Copilot Studio, Dataverse, Power Pages, Power Platform governance, ALM, DLP policies, CoE toolkit, or invokes /powerplatform-architect. Designs Power Platform solutions validated against the Power Platform Well-Architected Framework (5 pillars). Fetches latest documentation from Microsoft Learn MCP. Produces Dataverse data models, environment strategies, and ALM pipelines. DO NOT TRIGGER for Azure PaaS only (use azure-architect), D365 only (use d365-architect), or data analytics only (use data-architect).

Structured requirements gathering and fit-gap analysis for Microsoft architecture projects. TRIGGER when: user needs to elicit requirements, perform fit-gap analysis, capability mapping, MOSCOW prioritization, or invokes /requirements. Gathers functional and non-functional requirements, classifies fit-gap (OOB/Config/Custom/3rd Party), applies MOSCOW prioritization, and outputs a requirements register. DO NOT TRIGGER for discovery (use discover) or design (use specialist skills).

Security architecture specialist. TRIGGER when: user mentions Defender for Cloud, Microsoft Cloud Security Benchmark, Key Vault patterns, secret scanning, supply chain security, SBOM, private endpoints, NSG/ASG, WAF rules, OWASP, branch protection, or invokes /security-architect. Codifies opinions: Defender baseline non-negotiable, Key Vault references over inline secrets, RBAC over access policies, SBOM in CI, private endpoints in production. Reads from standards/references/security/stride-a-worksheet.md. DO NOT TRIGGER for identity (use identity-architect), deep Defender/Sentinel workflows (use defender-sentinel in Phase 4), or compliance audits (use threat-model in Phase 2).

Microsoft Tech-Design-First Specification Generator for epics and user stories. TRIGGER when: user asks to create a spec, generate a tech spec, write a design spec, create a feature spec, generate spec/design files, produce implementation specs for a user story or epic, or invokes /spec. Also triggers when user says "spec this", "design this feature", "create spec for epic", "generate design.md", "spec/design", or needs technical specifications for coding agents. Generates validated tech-design-first specifications with Mermaid LLD process diagrams, component interaction flows, and implementation tasks. DO NOT TRIGGER for full engagement workflow (use agent) or deep design review (use odin).

Site Reliability Engineering architect. TRIGGER when: user states an SLA, SLO, or availability target; user needs error budget governance, burn-rate alerting, chaos engineering, postmortem process, on-call rotation design, runbook authoring, incident command structure, or toil reduction strategy; user invokes /sre-architect. Codifies opinions: SLI/SLO/SLA derived in order (SLA from SLO, never invented separately), error budgets treated as real budgets (stop features when burned), multi-window multi-burn-rate alerting (Google SRE Workbook), blameless postmortems with owned action items, chaos in production with safety controls, runbooks retested every 90 days, toil below 50 percent of SRE time. Reads from standards/references/coding-stack/preferred-stack.md, standards/references/patterns/cloud-design-patterns.md, and standards/references/quality/review-checklist.md. DO NOT TRIGGER for App Insights or Log Analytics instrumentation setup (use observability-architect), Sentinel SOC analytics rules (use defender-s

Microsoft technology stack selection gate. TRIGGER when: user needs to choose between Microsoft technology stacks, asks about Stack A/B/C/D, needs help deciding between Power Platform, Azure PaaS, Containers, or Dynamics 365, or invokes /stack-select. Presents the 4-stack decision tree, gathers composition questions, and outputs a stack decision with ADR. This is a gate (no design work proceeds until a stack is selected). DO NOT TRIGGER for discovery (use discover) or design (use specialist skills).

Canonical standards reference for the Microsoft Architecture skill suite. Contains the preferred coding stack, functional programming paradigm, DDD principles, security-by-design checklists, review checklists, and C4 diagram guides. This skill is a reference library: other skills Read its references/ directory. DO NOT TRIGGER directly. Loaded by specialist skills on demand.

Threat modeling specialist. TRIGGER when: user mentions compliance, regulated workloads, PII, GDPR, CCPA, HIPAA, STRIDE, threat model, data breach, security risk assessment, OWASP for AI, prompt injection, MITRE ATT&CK, MITRE ATLAS, attack tree, blast radius, or when any vertical architect produces a security-sensitive design involving external-facing APIs, AI agents, or sensitive data flows. Also triggers when security-architect flags "formal STRIDE-A worksheet for regulated workloads" in its handoff. DO NOT TRIGGER for general security posture (use security-architect), identity primitives (use identity-architect), deep Defender/Sentinel SOC workflows (future defender-sentinel skill in Phase 4), or IaC delivery of controls (use iac-architect).

Target Operating Model (TOM) Architect for enterprise transformation. TRIGGER when: user asks to design a Target Operating Model, TOM, operating model, business transformation, process decomposition, capability map, maturity assessment, organizational design, service delivery model, or invokes /tom-architect. Also triggers for process taxonomy, L1 to L4 processes, capability register, maturity model, role mapping, KPI framework, AI augmentation overlay, service management framework, or when the problem requires decomposing business operations into structured layers. Covers Finance, HR, Procurement, Supply Chain, Cyber, and Sustainability domains. Maps TOM capabilities to Microsoft D365, Power Platform, Azure, and Fabric. Produces Mermaid diagrams, PPTX decks, and XLSX workbooks. DO NOT TRIGGER for pure technology architecture without business operating model context (use azure-architect, d365-architect, or other stack specialists). DO NOT TRIGGER for code-level specs (use spec).

Quality gate for Microsoft architecture deliverables. TRIGGER when: user wants to validate architecture artifacts, run quality checks, verify traceability, or invokes /validate. Dispatches three independent auditor subagents (Completeness, Consistency, Adversarial) in parallel, then synthesizes findings into a single Validation Report. DO NOT TRIGGER for discovery, design, or artifact generation: use the appropriate specialist skill.

Azure DevOps SAFe Backlog Exporter: transforms platform-neutral backlog into ADO-ready Excel workbook for CSV import. TRIGGER when: user asks to export backlog to Azure DevOps, create ADO work items, generate ADO import workbook, create SAFe backlog workbook, or invokes /backlog-ado. Also triggers for: "import to Azure DevOps", "ADO CSV import", "SAFe backlog Excel", "ADO work items from backlog". Maps 7 work item types (Epic, Feature, User Story, Technical User Story, Risk, Impediment, CI Item) to Azure DevOps SAFe field schema and generates import-ready Excel workbook. DO NOT TRIGGER for backlog decomposition (use backlog-decompose). DO NOT TRIGGER for Linear export (use backlog-linear) or GitHub export (use backlog-github).

PRD-to-Backlog Decomposition: platform-neutral hierarchical work item extraction. TRIGGER when: user asks to decompose a PRD into a backlog, extract work items from requirements, break down a PRD, create a backlog hierarchy, or invokes /backlog-decompose. Also triggers for: "backlog from PRD", "decompose requirements", "extract epics and stories", "PRD to work items". Produces platform-neutral hierarchical backlog with epics, features, user stories, technical stories, risks, impediments, and CI items. After decomposition, suggests the appropriate exporter skill for the user's target platform. DO NOT TRIGGER for platform-specific export (use backlog-ado, backlog-linear, or backlog-github). DO NOT TRIGGER for PRD creation (use prd-draft) or PRD review (use prd-review).

GitHub Issues Backlog Exporter: creates GitHub issues from platform-neutral backlog via MCP. TRIGGER when: user asks to create GitHub issues from backlog, export backlog to GitHub, push work items to GitHub Issues, or invokes /backlog-github. Also triggers for: "create GitHub issues", "push to GitHub Issues", "GitHub backlog import", "issues from backlog in GitHub". Creates hierarchical issues with labels, milestones, issue types, and sub-issue hierarchy via GitHub MCP server. DO NOT TRIGGER for backlog decomposition (use backlog-decompose). DO NOT TRIGGER for Azure DevOps export (use backlog-ado) or Linear export (use backlog-linear).

Linear Backlog Exporter: creates Linear issues from platform-neutral backlog via MCP. TRIGGER when: user asks to create Linear issues from backlog, export backlog to Linear, push work items to Linear, or invokes /backlog-linear. Also triggers for: "create Linear issues", "push to Linear", "Linear backlog import", "issues in Linear from backlog". Creates hierarchical issues with labels, projects, cycles, and parent-child relations via Linear MCP server. DO NOT TRIGGER for backlog decomposition (use backlog-decompose). DO NOT TRIGGER for Azure DevOps export (use backlog-ado) or GitHub export (use backlog-github).

Business Problem Discovery & Analysis for Product and Program Managers. TRIGGER when: user asks about business problem, problem discovery, understand this problem, business analysis, stakeholder mapping, pain point analysis, initiative classification, or invokes /discover. Also triggers when the problem requires structured discovery before solution design. Supports two entry modes: SaaS product discovery (Mode A) and consulting transformation discovery (Mode B). Produces a structured analysis document with problem statement, stakeholders, classification, and root causes. DO NOT TRIGGER for process flow mapping or persona mapping (use map). DO NOT TRIGGER for PRD generation (use prd-draft). DO NOT TRIGGER for PRD review (use prd-review). DO NOT TRIGGER for technical architecture (use enterprise-architect). DO NOT TRIGGER for Target Operating Model design (use tom-architect).

Epic Decomposition from upstream artifacts or direct input. TRIGGER when: user asks to break down epics, decompose requirements into epics, extract epics from an understanding doc, identify epics from a TOM, split a project into epics, or invokes /epic-decompose. Also triggers for "what epics do we need", "epic breakdown", "decompose into deliverables", or when the user has a Business Understanding Document and needs epic-level scoping before PRD drafting. DO NOT TRIGGER for PRD drafting or user stories (use prd-draft). DO NOT TRIGGER for PRD validation (use prd-validate). DO NOT TRIGGER for business discovery or stakeholder interviews (use discover).

Persona Mapping, Process Flows & Business Understanding Document assembly. TRIGGER when: user asks about persona mapping, process flow diagrams, user journey mapping, swimlane diagrams, BPMN flows, state machine diagrams, business understanding document, or invokes /map. Also triggers when discover hands off a completed analysis file. Produces a Business Understanding Document with embedded Mermaid process flows and persona profiles. DO NOT TRIGGER for problem discovery or classification (use discover). DO NOT TRIGGER for PRD generation (use prd-draft). DO NOT TRIGGER for technical architecture (use enterprise-architect). DO NOT TRIGGER for Target Operating Model design (use tom-architect).

Product Constitution: the living document that defines product principles, value propositions, positioning, customer experience philosophy, building approach, prioritization framework, and research bets. TRIGGER when: user asks about product principles, product philosophy, product constitution, "what do we stand for", product values, product positioning, prioritization framework, product strategy, product bets, "how do we decide what to build", or invokes /philosophy. Also triggers for "define our product principles", "update our positioning", "review our priorities", "what's our product philosophy". Supports three modes: Create (new constitution), Co-author (evolve sections), and Review (quarterly assessment). DO NOT TRIGGER for business discovery (use discover). DO NOT TRIGGER for PRD generation (use prd-draft). DO NOT TRIGGER for technical architecture (use tom-architect).