ravenclaude

Domain-neutral Team Lead + specialist-agents pattern for Claude Code: 15 specialists, 48 skills, gates, hooks, rules, templates. Capability Grounding Protocol, Structured Output Protocol, Researcher meta-skill, comfort-posture dashboard, opt-in command-review tribunal (the Thing), WebFetch sanitizer, branch-archive escape hatch, brand-extraction (homepage → brand kit), and declarative-visualization (Vega-Lite/Deneb/SVG, Gate 101 + Gate 103 svg-report-lint, viz-spec-reviewer). Slash commands: /init-agent-ready, /wrap, /set-posture, /dashboard, /forge, /reset-plugin-cache.

Power Platform specialist team — 11 agents (incl. power-platform-tester, power-bi-engineer) and 23 skills, with strong ALM/git coverage for solutions, flows, and PBIP. A house-opinions hook flags 8 §3/§4 violations; the knowledge bank carries production decision trees (PA-flow recovery, Dataverse token-acquisition, PCF React surface, PBI deploy/refresh, custom-connector build-path, PBIR Enhanced infinite-spinner debug + full build reference, DAX silent-zero scoring via the `Domain` pattern, sempy.fabric notebook reference, Power BI Copilot report-readiness, Code Apps connector gotchas, PBIP deployment variables + #839, PBIR reference enrichment, PBIP report fast-solve triage router [MCP-optional]) plus a real-engagement scenarios bank. Bundles the community pbix-mcp server (d0nk3yhm/pbix-mcp, MIT) for .pbix/.pbit read/write/DAX-eval (`pip install pbix-mcp`); documents (not bundles) the official Microsoft Dataverse MCP (CLAUDE.md §9a). Extends ravenclaude-core.

Corporate finance & FP&A specialist team — FP&A analyst, financial modeler, controller, treasury analyst, valuation analyst, audit-prep specialist, and board-pack composer. Ships 23 skills including the controller-autopilot: a governed close-to-report cycle (GAAP statements, COA mapping, reconciliation auto-match, review→approve→lock workflow, ELT staging, consolidation, per-entity dashboard, close schedules) plus its live-integration tier (multi-currency remeasurement + CTA, OAuth GL connectors + drill-through lineage, warehouse/RLS dashboard, IdP-backed segregation of duties) and a gold-standard NetSuite close (OAuth2 M2M + SuiteQL BS/IS trial balance, COA-draft, tie-out doctor, changed-after-sign-off drift, layperson runbook). 10 templates, a knowledge bank, and 2 advisory hooks (anti-patterns + secrets/PII scan). Inherits ravenclaude-core protocols (Grounding, Structured Output, Cited-Adjudicator).

Financial-regulatory & compliance specialist team — AML/KYC analyst (sanctions, EDD, SAR/STR), regulatory-reporting analyst (FATCA/CRS, supervisory returns, Solvency II, BMA EBS), risk-and-controls specialist (three lines of defense, KRIs, risk registers), policy & procedure writer, examination-prep specialist, and Bermuda-insurance specialist (BMA insurance code, captives, SAC, IL/ILS). Ships 4 skills (AML program review, regulatory mapping, SAR narrative drafting, examination readiness), 8 templates, and a defensive PII-scrubbing hook (SSN, IBAN, account numbers, Bermuda TIN patterns). Inherits ravenclaude-core protocols.

Web design & build specialist team — web architect (IA, stack, hosting), UX designer (wireframes, flows, conversion), visual designer (brand, typography, tokens), frontend implementer (HTML/CSS/JS, React, Astro, Next), content strategist (copy, microcopy, SEO content), accessibility auditor (WCAG 2.2 AA/AAA), performance engineer (Core Web Vitals, image/font optimization). 4 skills (design-system audit, accessibility review, Core Web Vitals tuning, technical SEO audit), 8 templates, 1 advisory hook, and a knowledge bank seeded with a curated 2026 reference set of "cutting edge yet simple" marketing-site donors (Linear, Vercel, Raycast, Resend, Cursor, v0, Tldraw, Cal.com) — visual-designer / ux-designer / frontend-implementer / web-architect all carry compact inline priors plus a pointer to the full brief. Inherits ravenclaude-core protocols.

EdTech-flavored Partner Success Manager team — 6 specialist agents anchored on the PSM lane with 8 skills, 15 templates, 1 advisory hook, and a 15-doc knowledge bank across 5 waves (v0.2.0 production-lesson + v0.3.0 foundational + v0.4.0 vertical-depth + v0.4.1 operating-cadence + v0.4.2 capability-extension). v0.4.2 adds 4 new skills (advocacy-program-design, adoption-sequencing-k12, implementation-90-day-arc, partner-training-program-design), 5 new knowledge files (edtech-reference-customer-patterns, k12-adoption-arc-fall-spring-summer, sis-sso-rostering-integration-patterns, district-implementation-failure-modes, k12-pd-norms-and-constraints), and 6 new templates (case-study-draft, reference-pipeline-tracker, adoption-diagnostic-worksheet, implementation-90-day-plan, train-the-trainer-curriculum, renewal-decision-memo). What would have been v0.5.0 team expansion, expressed as skills + knowledge + templates rather than parallel agents per the marketplace house rule (added 2026-05-21). Compact inline priors on the 6 affected agents; primary-source citations with retrieval dates. Vertical-explicit but segment-agnostic — fits K-12, higher-ed, corporate L&D, or mixed EdTech books. Inherits ravenclaude-core protocols including the alternate-methods Capability Grounding Protocol.

Data-platform team — 4 specialist agents (database-setup-guide, etl-pipeline-engineer, dashboard-builder, connector-developer) that guide cloud-database setup (Supabase / Fabric / RDS / Neon), ELT pipelines (Airbyte / Fivetran / n8n) with connectors for QuickBooks / Stripe / Salesforce / HubSpot / GA4 / Shopify / HRIS, and interactive dashboards (Evidence.dev / Apache Superset / Metabase / Cube + Tremor) embedded in marketing sites and client apps. 7 skills — three (jwt-embed-issuance, rls-policy-authoring, embed-csp-and-iframe-sandboxing) invoked by ravenclaude-core/security-reviewer; one (stack-selection) invoked by ravenclaude-core/architect. 12 templates (3 runnable security-critical + 4 conceptual + 3 seam-marked stubs + dbt-project-starter directory + cube-schema-starter with mandatory access_policy). 1 consolidated advisory hook. 13-doc knowledge bank: 7 v0.1.0 landscape + RLS + QBO + Power BI Embedded + EdTech LMS gap, plus 6 v0.2.0 per-connector deep-dives (Stripe, Salesforce, HubSpot, GA4, Shopify, HRIS). Opinionated against per-viewer-priced BI tools (Looker / Tableau / Sigma) for SMB-consulting use cases.

End-user authentication & identity team — 2 agents (auth-architect, auth-implementation-engineer) for adding login to a web app (React/Next), an API, and a dashboard. Methods: Google/Apple/Microsoft/GitHub social SSO, magic link, passkeys/WebAuthn, email+password — via managed auth (leaning Supabase Auth). 7 skills, 4 templates, a best-practices index, a scenarios bank (4 unverified field notes), an advisory anti-pattern hook, a stdlib JWT/scope/cookie analyzer (scripts/auth_analyze.py), and a web-verified knowledge bank with Mermaid decision trees (build-vs-buy, which-providers, OAuth-flow, token-storage, gate-the-dashboard, logout, refresh-reuse, MFA-factor, step-up auth). House rules: Authorization Code + PKCE never Implicit; never store tokens in localStorage; validate ID tokens server-side. Boundary: AUTHENTICATES the person; data-platform AUTHORIZES the data (RLS/embed-JWT); Entra seams to azure-cloud, auth code to security-reviewer. Requires ravenclaude-core@>=0.7.0.

Claude app-engineering specialist team — 6 agents (claude-solution-architect, prompt-and-context-engineer, mcp-and-server-tools-engineer, agent-sdk-engineer, eval-engineer, claude-app-ops-engineer) for building production applications on the Claude API, the Claude Agent SDK, and MCP. Covers the build-surface decision (Messages API / Agent SDK / Managed Agents), model right-sizing (Opus 4.8 / Sonnet 4.6 / Haiku 4.5, 1M context), prompt caching, tool use + structured output, MCP server authoring + Anthropic-hosted server tools (computer use, code execution, Files API, memory), evals (golden sets, LLM-as-judge), and FinOps + reliability (batch, model routing, backoff, observability). 15-doc dated-and-cited knowledge bank, 6 templates, 1 advisory hook (14 house opinions). Seams: prompt-as-artifact, security, and cross-domain architecture escalate to ravenclaude-core. Requires ravenclaude-core@>=0.7.0.

Cross-tool AI-coding model guidance — three strategist agents that help you reason about model selection in the non-Claude coding ecosystems: GitHub Copilot's model picker (completions / chat / coding agent / cloud agent + org model rules), OpenAI Codex (CLI + cloud model lineup, reasoning levels, /model), and xAI Grok (Grok 4.x lineup + the coding-tool surface). Teaches the DURABLE reasoning — task-complexity → reasoning-level → cost-tier → surface — over a single citation-grounded, dated lineup that is the marketplace's freshness anchor for non-Claude model facts. Volatile specifics (prices, context windows, picker availability) carry retrieval dates + a re-verify-at-use rider, and the agents fetch live before quoting. Seam: when the answer is a CLAUDE model or a Claude-app build, escalate to claude-app-engineering/claude-solution-architect — this plugin owns the non-Claude tools. Requires ravenclaude-core@>=0.7.0.

Project & delivery management team — four specialists across the predictive (PMBOK/PMP) and agile (Scrum/Kanban) tracks and the hybrid between: delivery-lead (charter, schedule, scope/change control, earned value), scrum-master (backlog, sprints, ceremonies, velocity, impediments), risk-and-raid-analyst (scored qual+quant risk, RAID depth, mitigation/contingency, issue triage), and stakeholder-comms-lead (stakeholder register, comms plan, status/exec reporting, escalation memos, steering packs). Deepens — does NOT replace — ravenclaude-core's domain-neutral project-manager (the lightweight RAID/status-hygiene default every plugin routes to); this is the deep PM craft layer. Knowledge: a predictive-vs-agile-vs-hybrid decision tree + a best-practices library. Seams: prose polish → ravenclaude-core/documentarian; system design → architect. Requires ravenclaude-core@>=0.7.0.

Database-engineering team — 4 agents (schema-architect, query-performance-engineer, migration-engineer, db-reliability-engineer) for the OPERATIONAL/transactional data layer (OLTP): relational schema design and normalization (and when to denormalize), indexing strategy and query/EXPLAIN-plan tuning, safe zero-downtime schema migrations (expand/contract), and connection-pooling, transactions/isolation, replication, and backup/restore reliability. Postgres-leaning, principles portable to MySQL/SQL Server; NoSQL access-pattern guidance where it fits. 5 skills, a decision-tree knowledge bank (index-choice + migration-safety trees + a dated 2026 capability map), 12 best-practices, 4 templates, 4 commands, 1 advisory hook. Distinct from data-platform (analytics/ELT/warehouse) and analytics-engineering (dbt). Seams: ELT/warehouse -> data-platform, ORM/app data-access -> backend-engineering, managed DB infra -> cloud plugins. Requires ravenclaude-core@>=0.7.0.

Backend-engineering team — 4 agents (backend-architect, service-implementation-engineer, backend-data-access-engineer, backend-reliability-engineer) for the application/service layer behind an API: domain modeling and service boundaries (monolith vs services, where to split), business-logic implementation and error handling, the data-access layer (repository pattern, ORM use, transaction boundaries, killing N+1), caching strategy (cache-aside, invalidation, stampede), background jobs and async messaging (idempotent workers, outbox, DLQs), and backend reliability (timeouts, retries with backoff + jitter, circuit breakers, graceful degradation). Language-agnostic (examples in Node/Python/Go). 6 skills, a decision-tree knowledge bank (service-boundary + cache trees + a dated map), 12 best-practices, 4 templates, 4 commands, 1 advisory hook. Seams: the contract -> api-engineering, the schema/index -> database-engineering, deploy -> devops-cicd. Requires ravenclaude-core@>=0.7.0.

Mobile-engineering team — 4 agents (mobile-architect, ios-engineer, android-engineer, cross-platform-engineer) for native and cross-platform mobile apps: the native-vs-React-Native-vs-Flutter decision, iOS craft (Swift/SwiftUI, the app lifecycle, concurrency), Android craft (Kotlin/Jetpack Compose, lifecycle, coroutines), and the cross-cutting mobile concerns (offline-first + sync, secure on-device storage, push notifications/deep links, battery/network constraints, and the App Store / Play Store release pipeline + review). 5 skills, a decision-tree knowledge bank (platform-choice + offline-sync trees + a dated 2026 capability map), 12 best-practices, 4 templates, 4 commands, 1 advisory hook. Seams: shared web patterns -> frontend-engineering, the backend API -> api-engineering/backend-engineering, auth -> auth-identity, CI/CD signing -> devops-cicd. Requires ravenclaude-core@>=0.7.0.

Google Cloud (GCP) infrastructure & platform team — 4 agents (gcp-architect, gcp-iam-engineer, gcp-network-engineer, gcp-data-and-compute-engineer) for the GCP layer of a stack: the resource hierarchy (organization/folders/projects) and org policy constraints, least-privilege IAM (predefined over primitive roles, service accounts + Workload Identity Federation, no key files), VPC networking (Shared VPC, firewall rules, Private Service Connect, Cloud NAT), compute selection (Cloud Run / GKE / Cloud Functions / GCE), and the data services (BigQuery, Pub/Sub, Cloud SQL) at a selection level. 5 skills, a decision-tree knowledge bank (compute-selection + project-hierarchy trees + a dated 2026 capability map), 12 best-practices, 4 templates, 4 commands, 1 advisory hook. Multi-cloud seam to aws-cloud/azure-cloud; IaC -> terraform-iac; BigQuery analytics -> data-platform; security verdicts -> ravenclaude-core/security-reviewer. Requires ravenclaude-core@>=0.7.0.

Technical-writing & docs team — 3 agents (docs-architect, api-reference-writer, docs-site-engineer) for developer and product documentation that people actually use: the Diataxis framework (tutorial / how-to / reference / explanation — and not mixing them), docs-as-code (docs in the repo, reviewed and CI-checked, versioned with the product), accurate API/SDK reference (spec-driven, examples that run, errors documented), and a maintainable docs site (Docusaurus/Mintlify/MkDocs, IA, search, versioning). Deepens ravenclaude-core/documentarian. 5 skills, a decision-tree knowledge bank (doc-type, docs-tooling, Diataxis content-type, lint-in-CI-vs-manual + a dated 2026 map), a 4-scenario bank, 12 best-practices, 4 templates, 4 commands, 1 advisory hook, a prose-linter LSP tier (Vale-LS + Marksman) and a stdlib readability/prose checker. Seams: dev portal/SDK -> api-engineering, site design -> web-design, prose polish -> ravenclaude-core/documentarian. Requires ravenclaude-core@>=0.7.0.

Product-management team — 3 agents (product-strategist, product-discovery-lead, product-metrics-analyst) for deciding WHAT to build and WHY: product strategy and vision (positioning, the strategy stack, opportunity sizing), continuous discovery (customer interviews, opportunity-solution trees, assumption testing, JTBD), evidence-based prioritization (RICE/cost-of-delay over opinion), crisp PRDs/specs framed as problems + outcomes, and product metrics (a North Star + input metrics, not vanity). Distinct from project-management (delivery/schedule/RAID) — this is the what/why, that's the how/when. 5 skills, a decision-tree knowledge bank (build-vs-not + prioritization + build-vs-buy-vs-partner + a dated 2026 map), a scenarios bank, pm_calc.py (RICE/WSJF/opportunity), 12 best-practices, 4 templates, 4 commands, 1 advisory hook. Seams: delivery -> project-management, experiments -> applied-statistics, UX -> web-design, A/B infra -> experimentation-growth-engineering. Requires ravenclaude-core@>=0.7.0.

Experimentation & growth-engineering team — 3 agents (experimentation-architect, feature-flag-engineer, product-analytics-instrumentation-engineer) for the APPARATUS that lets a team learn fast and ship safely: feature flags + progressive delivery (kill switches, targeting, flag lifecycle/debt), A/B-test plumbing (assignment, exposure logging, SRM checks, sequential-peeking guards, the results read), and trustworthy event instrumentation (a tracking plan, a consistent event schema, identity stitching, a CDP). The engineering of experimentation — statistical VALIDITY (power/MDE, significance) is applied-statistics'. 5 skills, a decision-tree knowledge bank (flag-vs-config + ship-decision trees + a dated 2026 map), 12 best-practices, 4 templates, 4 commands, 1 advisory hook. Seams: significance -> applied-statistics, hypothesis/outcome -> product-management, rollout -> devops-cicd, pipelines -> data-platform. Requires ravenclaude-core@>=0.7.0.

Analytics-engineering team — 3 agents (analytics-engineer, semantic-layer-engineer, data-quality-testing-engineer) for the TRANSFORMATION layer in the modern data stack: dbt modeling (staging -> intermediate -> marts, the medallion/Kimball split, incremental models, materializations), a governed semantic/metrics layer (one definition of revenue/active-user, metrics-as-code) so every tool agrees, and data quality (dbt tests, freshness, model contracts, anomaly checks) that gates the warehouse. Warehouse-neutral (Snowflake/BigQuery/Redshift/Databricks). 5 skills, a decision-tree knowledge bank (materialization + model-layer trees + a dated 2026 map), 12 best-practices, 4 templates, 4 commands, 1 advisory hook. Distinct from data-platform (ingestion/warehouse/BI) and database-engineering (OLTP). Seams: ELT/warehouse provisioning -> data-platform, OLTP -> database-engineering, BI -> tableau/data-platform, enterprise lakehouse -> microsoft-fabric. Requires ravenclaude-core@>=0.7.0.

ML-engineering (MLOps) team — 4 agents (ml-platform-architect, training-pipeline-engineer, model-serving-engineer, ml-monitoring-engineer) for the PRODUCTION lifecycle of ML models: the platform/architecture (build-vs-buy, the stack, the train->register->serve->monitor loop), reproducible training pipelines + experiment tracking + a model registry, feature stores and train/serve consistency (avoiding training-serving skew and leakage), model serving (online vs batch, shadow/canary), monitoring (data + concept drift, decay, retraining triggers), and computer-vision MLOps (task->architecture + edge-vs-cloud inference). 6 skills, a decision-tree knowledge bank (serving-pattern + retraining + computer-vision trees + a dated 2026 map), 25 best-practices, 4 templates, 4 commands, 1 advisory hook. Seams: significance -> applied-statistics, data pipelines -> data-platform/data-streaming, LLM/agent apps -> claude-app-engineering, deploy -> devops-cicd/cloud-native-kubernetes. Requires ravenclaude-core@>=0.7.0.

Data governance & privacy team — 3 agents (data-governance-architect, privacy-compliance-engineer, data-catalog-lineage-engineer) for making data discoverable, trustworthy, and lawfully used: the governance operating model (ownership/stewardship, policies, classification scheme), privacy mechanics (GDPR/CCPA/-style data-subject rights — access/erasure/portability, consent + lawful basis, data minimization, retention, pseudonymization vs anonymization), and a data catalog with lineage + sensitive-data/PII discovery + access governance/DLP. The ENGINEERING of governance, not legal advice. 5 skills, a decision-tree knowledge bank (classification + DSR-handling trees + a dated 2026 map), 12 best-practices, 4 templates, 4 commands, 1 advisory hook. Seams: financial-regulatory -> regulatory-compliance, warehouse RLS -> data-platform, Purview-in-Fabric -> microsoft-fabric, security verdicts -> security-engineering. Requires ravenclaude-core@>=0.7.0.

Security-engineering (AppSec) team — 4 agents (appsec-engineer, threat-modeler, supply-chain-security-engineer, cloud-security-engineer) for building security INTO software: threat modeling (STRIDE, data-flow diagrams, trust boundaries), application security testing (SAST/DAST/IAST/SCA in CI, OWASP Top 10 web), secrets management (detection, rotation, vaulting, no plaintext), software supply-chain integrity (SBOM consumption, dependency CVEs, SLSA verification, pinning), and cloud security posture (CSPM, IAM least-privilege, network exposure). 5 skills, a decision-tree knowledge bank (vuln-triage + secrets-handling trees + a dated 2026 tooling map), 12 best-practices, 4 templates, 4 commands, 1 advisory hook. **Proposes controls; every ship/no-ship security VERDICT escalates to ravenclaude-core/security-reviewer.** Seams: API OWASP -> api-engineering, identity -> auth-identity, data privacy -> data-governance-privacy. Requires ravenclaude-core@>=0.7.0.

Cloud-native & Kubernetes team — 4 agents (kubernetes-architect, container-build-engineer, k8s-platform-operator, service-mesh-networking-engineer) for running workloads on Kubernetes well: workload design (Deployment/StatefulSet/Job, probes, requests/limits, PDBs, HPA/VPA), container build hygiene (minimal non-root images, distroless, multi-stage), platform ops (namespaces, RBAC, network policies, resource quotas, admission control, upgrades), and service-mesh networking (ingress/gateway, mTLS, traffic-splitting, observability). Cloud-agnostic. 6 skills, a decision-tree knowledge bank (workload-type + ingress trees + a dated 2026 ecosystem map), 12 best-practices, 4 templates, 4 commands, 1 advisory hook. Seams: managed control plane (AKS/EKS/GKE) -> azure/aws/gcp-cloud, GitOps reconcile -> devops-cicd, telemetry/SLOs -> observability-sre. Requires ravenclaude-core@>=0.7.0.

Terraform & IaC team — 3 agents (iac-architect, terraform-module-engineer, iac-policy-and-state-engineer) for cloud-agnostic infrastructure-as-code: module design (composable, versioned, registry-published, the root-vs-child split), remote state and backend safety (locking, isolation by blast radius, no secrets in state), environment promotion (workspaces vs directories vs Terragrunt), drift detection, and policy-as-code guardrails (OPA/Sentinel/Conftest) in the plan pipeline. Terraform + OpenTofu. 5 skills, a decision-tree knowledge bank (state-isolation + module-boundary trees + a dated 2026 tooling map), 12 best-practices, 4 templates, 4 commands, 1 advisory hook. Seams: per-cloud resources -> azure-cloud (Bicep too)/aws-cloud/gcp-cloud, plan in CI -> devops-cicd, posture verdicts -> security-engineering. Requires ravenclaude-core@>=0.7.0.

Hospice referral-sales / community-liaison team — 6 agents (referral-development-strategist, hospice-eligibility-educator, referral-account-manager, admissions-conversion-coach, goals-of-care-conversation-coach, hospice-sales-compliance-advisor) for a hospice sales / community-education rep who grows admissions by developing referral sources (hospitals, SNFs, physician practices, ACOs): territory planning, educating clinicians on hospice eligibility (LCD decline criteria by diagnosis — the rep educates, the physician certifies), referral-partner reviews, the referral-to-admission funnel, the goals-of-care 'hospice conversation', and clearing every touch against the Anti-Kickback / HIPAA line. 6 skills, 6 commands, a 4-doc knowledge bank (6 decision trees + glossary + LCD + compliance references), 14 best-practices, scenarios, and a runnable funnel / census / benefit-period calculator. Employer-neutral; public practice + CMS rules. Requires ravenclaude-core@>=0.7.0.

AI/LLM red-teaming team — 2 agents (ai-redteam-lead, adversarial-testing-engineer) for the layer answering 'can this AI system be made to do harm, leak data, or exceed its authority — and how do we harden it?': threat modeling + rules of engagement, the attack taxonomy (OWASP LLM Top 10 2025 + MITRE ATLAS), direct vs indirect prompt injection, jailbreaks (roleplay/encoding/many-shot/crescendo), data exfiltration & training-data extraction, agentic tool-abuse / excessive agency, multimodal attacks, and defense-in-depth remediation. Fluent in automated red-team harnesses (PyRIT, Garak, Promptfoo red-team, Giskard) and likelihood×impact severity. 3 skills, a 2-doc knowledge bank (attack-taxonomy decision tree + 2026 patterns), and 2 templates. Distinct from llm-evaluation-engineering (quality-regression eval), trust-and-safety (content-moderation / T&S policy), and security-engineering (app/infra pentest) — the adversarial AI-security layer over model- and agent-based systems. Requires ravenclaude-core@>=0.7.0.

Chaos-engineering & resilience team — 2 agents that make distributed systems survive failure. resilience-architect owns the DESIGN side: failure-mode analysis (FMEA), resilience patterns (timeouts, retries with backoff+jitter, bulkheads, circuit breakers, load shedding, graceful degradation, idempotency), the observability-maturity gate for whether you're ready to run chaos, and redundancy/DR (multi-AZ/region, failover, RTO/RPO). chaos-experiment-engineer owns the EXPERIMENT + VERIFICATION side: hypothesis-driven experiments, blast-radius containment (automatic abort/rollback), game days, fault injection (latency/error/resource/dependency-outage/partition/zone), and verifying the pattern held under load+fault. 3 skills, a 2-doc knowledge bank (decision trees + a dated 2026 patterns reference), 2 templates. Distinct from observability-sre (metrics/SLO/on-call — a hard prerequisite), devops-cicd (progressive delivery), performance-engineering (load), incident-response-dfir. Requires ravenclaude-core@>=0.7.0.

Serverless engineering team — 2 agents for building on serverless the provider-neutral way. serverless-architect owns the design side: event-driven decomposition, FaaS-vs-containers-vs-managed-services and when serverless is the WRONG choice, sync/async, orchestration-vs-choreography, event contracts, the dual-write/outbox problem, and serverless-friendly storage. serverless-runtime-and-ops-engineer owns the runtime side: cold starts, concurrency & throttling, idempotency & exactly-once-effect, DLQs & poison messages, retries, per-invocation cost & the scale crossover, tracing in ephemeral runtimes, and local emulation. 3 skills, a 2-doc knowledge bank (a Mermaid decision tree + a dated 2026 patterns reference), 2 templates. Provider-neutral PATTERNS only — distinct from aws-cloud/gcp-cloud/azure-cloud (provider IaC/specifics), data-streaming-engineering (Kafka/Kinesis), devops-cicd (deploy pipeline), backend-engineering (the serverful alternative), observability-sre (SLOs). Requires ravenclaude-core@>=0.7.0.

UX-research team — 2 agents (ux-research-lead, research-execution-and-synthesis-specialist) that turn a fuzzy ask into a researchable question, pick a rigorous method (generative vs evaluative, qual vs quant, moderated vs unmoderated), run the study (usability tests, interviews, surveys), and synthesize evidence into decision-ready, bias-checked insight — without the research theater. The lead owns framing, method selection, the research plan, research-ops (recruiting, repository), and research ethics + participant-PII discipline (consent, anonymization). The specialist owns protocol & task design, non-leading moderation, survey design, and synthesis (affinity/thematic analysis, evidence strength, observation vs interpretation). 3 skills, a 2-doc knowledge bank + 2 templates. Distinct from product-management (roadmap/prioritization), web-design (visual/interaction/IA design), experimentation-growth-engineering (online A/B stats), data-science-research (advanced quant). Requires ravenclaude-core@>=0.7.0.

Standalone brand-CREATION service for a web studio's commission customers — 2 agents (brand-strategist, identity-systems-designer): strategy/positioning/archetype, naming & tagline, curated logo suite, color/type systems (WCAG-paired), brand book, legal/IP posture, collateral, sellable tiers. AI-drafts + MANDATORY human curation. Thin orchestration layer: delegates design tokens to web-design:design-tokens-scaffolding, raw generation + license/indemnity to generative-web-media, and hands the finished system to web-design:visual-designer. 5 skills, 4 commands, a 3-doc knowledge bank (Mermaid trees), 5 best-practices, 6 templates, a stdlib WCAG-pair checker, 1 advisory hook. Not legal advice — every IP / font-license / registrability claim routes to security-reviewer. Prices [unverified]. Requires ravenclaude-core@>=0.7.0.

AI multimedia for brand & winery sites: a creative brief -> on-brand, web-optimized, license-clean images/video/3D/audio behind a mandatory human curation gate. 4 agents (generation-strategist, web-asset-pipeline-engineer, asset-provenance-guardian, brand-and-accessibility-reviewer) route a brief to the right generator (provider-neutral, Grok-lean for images where competitive; inpaint/outpaint/bg-removal/upscale first-class), turn raw output into AVIF/WebP responsive <picture> markup with LCP/CLS-safe embeds, pin commercial-use licenses (flags the FLUX-dev non-commercial trap; C2PA + a provenance ledger; EU AI Act Art.50 disclosure), and gate every asset on brand-hex/style conformance + WCAG alt text before ship. 6 skills, 4 knowledge docs (Mermaid decision trees; all prices [unverified]), 6 best-practices, 4 commands, 5 templates, 4 scripts. The shared foundation the brand-identity-studio plugin consumes. Declarative fal MCP binding (set FAL_KEY). Requires ravenclaude-core@>=0.7.0.

Web commerce integration specialist — scaffolds production-ready payment/commerce backends (Stripe, Square, or Shopify; one per site) into static or JS-framework sites at full depth: catalog, cart, hosted/embedded checkout, verified webhooks, idempotency, POS/inventory reconciliation, and order handling. Three first-class provider tracks behind a thin shared payment-lifecycle contract — NOT a leaky unified adapter. Generates first-party code (no dormant/deprecated deps), with PCI card-isolation, webhook signature verification, idempotency, and env-only secrets as GENERATED-CODE invariants. The backend-integration lane that plugs into web-design's build pipeline; every payment/PII security verdict routes to ravenclaude-core/security-reviewer. Inherits ravenclaude-core protocols.

Internal audit team — 2 agents (internal-audit-lead, audit-engagement-specialist) for the independent assurance & advisory function over ALL risk, anchored on the IIA Global Internal Audit Standards (2024 — 5 domains / 15 principles), COSO Internal Control & COSO ERM, and the Three Lines Model. Covers the risk-based audit universe → annual plan → engagement lifecycle (planning memo/scope, risk & control matrix, test of design + operating effectiveness, attribute sampling, workpapers & evidence), findings via the 5 C's, impact×likelihood issue rating, the audit-committee report & management action plans, follow-up/remediation validation, and the QAIP + external quality assessment. Independence discipline: IA assures/advises, never owns controls. 3 skills, a 2-doc knowledge bank (decision tree + 2026 patterns), and 2 templates. Distinct from cybersecurity-grc (security-control assurance), regulatory-compliance (AML/financial regs), and esg-sustainability-reporting (ESG assurance). Needs ravenclaude-core.

Corporate treasury team — 2 agents (treasury-strategy-lead, cash-and-risk-operations-specialist) answering 'how much cash, where — who banks us, and how do we protect the money?': cash & liquidity forecasts (CCC, direct/indirect, 13-week), liquidity-buffer & facility policy, bank-relationship & account structure (ZBA, notional/physical pooling, in-house bank, AFP fees, BAI/ISO 20022), investment policy (IPS) & debt covenants, FX & rate hedging (transaction/translation/economic exposure, forwards/swaps/options/collars, ASC 815 / IFRS 9 hedge accounting), payments & fraud controls (positive pay, dual auth, BEC), working-capital & TMS selection. 3 skills, a 2-doc knowledge bank, and 2 templates. Distinct from finance (FP&A/budgeting), fintech-payments-engineering (rail code), and regulatory-compliance (AML/OFAC) — treasury owns the CASH and the BANK RELATIONSHIP. Not legal/tax/accounting advice; volatile hedge-accounting & bank/regulatory specifics carry a retrieval date + verify-at-use. Needs ravenclaude-core.

Quantum computing engineering team — 2 agents (quantum-solutions-architect, quantum-algorithm-engineer) answering 'is this even a quantum problem, and if so — which paradigm, modality, and algorithm, and can we trust the result?': hard TRIAGE (most problems win classically today), gate-model vs annealing, qubit modalities (superconducting/trapped-ion/neutral-atom/photonic), NISQ vs fault-tolerant (logical vs physical qubits, surface code), transpilation, VQE/QAOA, error MITIGATION (ZNE/PEC) vs CORRECTION, hybrid loops, and QPU access via SDKs (Qiskit/Cirq/PennyLane/Braket). 3 skills, a 2-doc knowledge bank, and 2 templates. Engineering judgment, not a benchmark; the device/SDK/qubit-count landscape is volatile — every version carries a retrieval date + [verify-at-use]. Distinct from ml-engineering (classical MLOps), hardware-electronics-engineering (control board/cryo), security-engineering/cybersecurity-grc (post-quantum crypto defense), and performance-engineering (HPC). Needs ravenclaude-core.

Pricing & monetization team — 2 agents (pricing-strategist, monetization-analyst) that own the question no other plugin answers: what to charge and how to package it. Covers pricing-model selection (subscription / usage / per-seat / tiered / freemium / hybrid), the value-metric choice, packaging & tiering (good-better-best, fencing, add-ons), willingness-to-pay research (Van Westendorp, Gabor-Granger, conjoint), discount & price-change governance, and monetization metrics (ARPA, NRR, expansion, discount leakage, the rule-of-40 caveat). 5 skills, a 3-doc knowledge bank with Mermaid decision trees (model selection, value-metric, WTP-method) + a dated 2026 reference, 10 best-practices, 3 templates. Domain-neutral. Seams: model mechanics → finance; what-to-build → product-management; quota/comp → sales-revops; is-the-price-test-real → applied-statistics. Requires ravenclaude-core@>=0.7.0.

Prompt & context engineering team — 3 agents (prompt-architect, prompt-implementation-engineer, prompt-reliability-engineer) that own the layer between the model and the app: prompt & context strategy (task decomposition, pattern selection across few-shot / CoT / decomposition / role, system-vs-user structure), the output-format contract (JSON mode, tool/function-calling schemas, few-shot curation, token budgeting), and prompt reliability (eval sets & regression, prompt versioning/CI, prompt-injection & jailbreak defense at the prompt layer). Provider-neutral. 4 skills, a 2-doc knowledge bank with 4 Mermaid decision trees + a dated 2026 reference, 5 best-practices, 2 templates. Seams: which model → ai-coding-model-guidance / claude-api; retrieval → ai-rag-engineering; eval at scale → llm-evaluation-engineering; adversarial testing → ai-red-teaming; the app → claude-app-engineering / backend-engineering. Requires ravenclaude-core@>=0.7.0.

Database Reliability Engineering (DBRE) team — 3 agents (dbre-architect, database-operations-engineer, database-incident-responder) that own the PRODUCTION database as a reliability surface, the layer database-engineering (schema/query design) doesn't: HA topology & failover from RPO/RTO, backup & point-in-time recovery with restore verification, disaster recovery, capacity planning, connection pooling; zero-downtime expand-contract migrations, backfills, replication management & failover drills, upgrades; and DB on-call (replication lag, lock contention, connection storms, runaway queries, disk-full), DB SLOs, blameless postmortems. 4 skills, a 2-doc knowledge bank with 4 Mermaid decision trees + a dated 2026 reference, 5 best-practices, 2 templates. Seams: schema/query → database-engineering; service SRE → observability-sre; provisioning → terraform-iac / cloud; pipelines → data-orchestration; access → security-engineering / auth-identity. Requires ravenclaude-core@>=0.7.0.

Regenerates a review-ready report (HTML or Office/Word→PDF) from an old distributed report used as a template — infers the template's structure with no markup, re-binds it to new data (including Power BI data via XMLA, with a REST executeQueries fallback, and an embedded Power BI screenshot), and runs a gold-standard auto-QA gate: data accuracy vs source, template fidelity, WCAG/PDF-UA accessibility, consistency, and a blocking no-old-client-data-leak egress scan. It GUARANTEES only (a) no old-client-data leak and (b) that every low-confidence classification is surfaced for human review — it produces a draft; peer review, edits, and distribution stay with the human. Local-only for sensitive client data. Skills ship per lane (v0.1.0 HTML → v0.2.0 Office → v0.3.0 Power-BI-ingest); counts reconciled at Gate 12 PR time. Requires ravenclaude-core@>=0.204.0.