Communitygithub.com

Crapule421/tiger-team

Enterprise architecture knowledge base - design authority role, architecture principles, target vs transition architectures, macro integration patterns, non-functional requirements (availability, RPO/RTO, performance, security, operability, reversibility), cloud strategy and sovereignty (SecNumCloud, reversibility), technical debt, DORA ICT risk from an architecture view, and a cross-expert coherence checklist. Load when designing or reviewing a target architecture, arbitrating structural options, or writing ADRs; references cover TOGAF/ArchiMate, French urbanisation du SI and ADR formats.

What is tiger-team?

tiger-team is a Claude Code agent skill that enterprise architecture knowledge base - design authority role, architecture principles, target vs transition architectures, macro integration patterns, non-functional requirements (availability, RPO/RTO, performance, security, operability, reversibility), cloud strategy and sovereignty (SecNumCloud, reversibility), technical debt, DORA ICT risk from an architecture view, and a cross-expert coherence checklist. Load when designing or reviewing a target architecture, arbitrating structural options, or writing ADRs; references cover TOGAF/ArchiMate, French urbanisation du SI and ADR formats.

Works with✓Claude Code~Codex CLI~Cursor
npx skills add https://github.com/Crapule421/tiger-team/tree/HEAD/plugins/tiger-team/skills/architecture-reference

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Architecture Reference (Enterprise Architect / Design Authority)

For regulated EU/French contexts (banking, healthcare, defense). Regulatory details evolve: anything marked "verify" must be checked against the current official text before being used in a deliverable.

References index

  • references/togaf-archimate.md — read when structuring an architecture effort (ADM phases, deliverables, governance, compliance reviews) or modelling with ArchiMate (layers, elements, relationships, viewpoints). Full path ${CLAUDE_PLUGIN_ROOT}/skills/architecture-reference/references/togaf-archimate.md
  • references/urbanisation.md — read when producing or using French IS maps (zones/quartiers/îlots, cartographies), a schéma directeur, application portfolio assessment (TIME) or a rationalisation/decommissioning plan. Full path ${CLAUDE_PLUGIN_ROOT}/skills/architecture-reference/references/urbanisation.md
  • references/adr.md — read when writing or reviewing an Architecture Decision Record or a decision matrix (Nygard, MADR, Y-statements, weighted criteria, fictitious examples). Full path ${CLAUDE_PLUGIN_ROOT}/skills/architecture-reference/references/adr.md

Related deliverable templates: ${CLAUDE_PLUGIN_ROOT}/skills/deliverables/templates/dat.md, interface-contract.md, flow-matrix.md. Regulatory detail (GDPR, DORA, NIS2, HDS, II 901, IGI 1300…) lives in the sector-compliance skill and its sector references (${CLAUDE_PLUGIN_ROOT}/skills/sector-compliance/references/); this skill covers the architecture implications only.

1. Role: design authority

Owns: architecture principles and standards; target architecture and roadmap; structural decisions (ADRs); coherence across domains and experts; dispensations (dérogations) with expiry; architecture compliance reviews; technical debt register at architecture level. Does not own: security risk acceptance (business risk owner / homologation authority, advised by RSSI); personal data decisions (controller, advised by DPO); budget/schedule arbitration (COPIL); production changes (CAB).

Review gates

GateWhenChecks
G0 ScopingNote de cadrageFit with target/roadmap, options identified, regulatory perimeter, sovereignty class of data
G1 DesignDAT / interface contracts draftPrinciples compliance, NFRs quantified, integration pattern, data ownership, security architecture, reversibility
G2 Pre-buildDesign frozenADRs accepted, dispensations signed, flow matrix complete, ops/DR design
G3 Pre-go-liveBefore go/no-goAs-built vs DAT, debt incurred registered, DR tested, DORA register / dependency map updated

Outcome of a review: compliant / compliant with reservations (actions + dates) / dispensation (risk owner + expiry) / non-compliant (blocking).

2. Architecture principles

Format (TOGAF style): Name — Statement — Rationale — Implications. Keep 10-15, each testable in a review. Default set for regulated integration contexts:

#PrincipleTestable implication
P1Business continuity by designEvery critical function has stated RTO/RPO met by design and tested
P2Security and privacy by design and by default (GDPR Art. 25)Data classified (DICT) before design; least privilege; encryption in transit and at rest
P3Single source of truth per data domainEach data entity has one master system and a named owner; others hold governed copies
P4Loose coupling through managed interfacesNo direct access to another application's database; all exchanges via contracted APIs, messages or files
P5Reuse before buy before buildExisting shared platforms (MFT, API gateway, IAM, scheduling) used unless an ADR says otherwise
P6Standards and open formatsOpen protocols/formats preferred; proprietary only with reversibility plan
P7Reversibility and exitEvery external service has an exit plan; data exportable in documented formats
P8Operability is a requirementMonitoring, logs, scheduling, restart and DEX designed with the solution
P9Traceability and auditabilityEnd-to-end correlation IDs; audit trails retained per legal requirements
P10Data residency and sovereignty by classificationHosting choice derived from data class (see §6)
P11Simplicity and rationalisationNew component only if no existing one fits; decommissioning planned with every replacement
P12Automate the delivery chainInfrastructure and deployments as code, reproducible environments

3. Target vs transition architectures

  • Baseline (as-is), Target (to-be at horizon, typically 3-5 years), Transition architectures (plateaus: stable, operable intermediate states delivering value), Gap analysis (what is added, removed, changed), Roadmap (work packages, dependencies, decommissioning dates).
  • Each transition must be operable on its own (supported, monitored, DR-covered, documented) — the project may stop there.
  • Coexistence bridges (temporary sync, dual-write avoidance, anti-corruption layers) are first-class components: owner, cost, sunset date, debt entry.
  • "Temporary" without a sunset date becomes permanent: refuse it at G1.
  • Prefer incremental migration (strangler fig: route functionality progressively to the new system behind a façade) over big-bang replacement when the legacy is large and poorly documented.

4. Integration architecture (macro level)

Detailed design belongs to integration-expert; the architect sets the pattern and the rules.

PatternFitsAvoid whenWatch-outs
Point-to-point≤ few stable flows, tacticalMany consumers, frequent changeSpaghetti (n² links), no central monitoring
Hub & spoke / EAI / ESBMany heterogeneous systems, transformations, central governanceHub becomes bottleneck/monolith of business logicKeep business logic out of the bus
File transfer via MFTBatch, high volume, partner exchanges, legacy (very common in banking/IBM i)Real-time needsIntegrity checks, naming, cut-off times, re-send, encryption, non-repudiation
API-led (system / process / experience layers)Reusable services, digital channels, partner APIsBulk data movementAPI management, versioning, quotas, OAuth2/mTLS
Event-driven (pub/sub, event streaming, CDC)Decoupled reactions, real-time propagation, many consumersStrong transactional consistency across systems requiredOrdering, idempotency, schema evolution, replay, event ownership
Data platform (DWH, lake, lakehouse, data mesh)Analytics, regulatory reporting, MLOperational transactional integrationLineage, quality, retention, sensitive data segregation
MDM / référentielsShared master data (customer, product, organisation, patient identity)Data with a single natural ownerGovernance and stewardship effort is the real cost

Rules of thumb: synchronous calls only when the caller truly needs the answer now; async by default between domains; design every flow for failure (retry with backoff, dead-letter, replay, idempotent consumers); use the outbox pattern rather than dual writes; canonical data models only within a bounded scope (enterprise-wide canonical models rarely survive); version every contract; one correlation ID end-to-end.

5. Non-functional requirements

Quantify every NFR; "highly available" is not a requirement.

CategoryMetrics / questionsPitfalls
Availability% over a period, service hours, planned downtime allowedIgnoring dependency chain; measuring infra not service
ContinuityRTO / RPO (DMIA / PDMA) per business function, MTPDRPO 0 promised with async replication; batch restart ignored
PerformanceThroughput, p95/p99 latency, batch window, peak factor, growthTesting average not peak; month-end peaks forgotten
ScalabilityVolumes at horizon, elasticityLicence models that scale cost non-linearly
SecurityDICT rating, IAM, encryption, logging, segregationSecurity bolted on after design
OperabilityMonitoring, alerting, logs, restart, scheduling, automationDEX written after go-live
ObservabilityTraces, metrics, logs, correlation, business KPIsNo end-to-end view across middleware
ReversibilityExit plan, export formats, time and cost to exitOnly contractual, never tested
InteroperabilityStandards (e.g., ISO 20022 in payments, HL7 FHIR in health — verify sector mandates)Local dialects of standards
Compliance / residencyData location, retention, legal holds, auditBackups and logs stored outside the agreed zone
MaintainabilityUpgrade cadence, vendor support datesEnd of support hitting mid-project
SustainabilityResource efficiency (RGESN eco-design guidelines in France — verify status)Oversized non-production environments

Availability cheat sheet (per year): 99 % ≈ 3.65 days; 99.5 % ≈ 1.83 days; 99.9 % ≈ 8.76 h; 99.95 % ≈ 4.38 h; 99.99 % ≈ 52.6 min; 99.999 % ≈ 5.3 min. Serial chain: A = A1 × A2 × … ; redundant pair: A = 1 − (1 − A1)(1 − A2) (assumes independent failures — rarely fully true). RPO/RTO coherence: RPO is bounded by replication mode and backup frequency; RTO includes detection, decision, failover, data repair, batch catch-up and partner re-synchronisation — not only technical switch time.

6. Cloud strategy & sovereignty

Classify first, then host: the data class drives the hosting option.

Data / system classTypical hosting options (verify against current rules)
Public / internal, non-sensitivePublic cloud acceptable with standard controls
Personal data (GDPR)EU hosting preferred; transfers outside EEA need Chapter V basis and transfer impact assessment
Health personal data (France)HDS-certified hosting provider required (Code de la santé publique, Art. L.1111-8); recent HDS referential adds data-location requirements in the EEA — verify current version and transition dates
Data of particular sensitivity (données d'une sensibilité particulière) of State services, State operators and listed GIPSecNumCloud-qualified offer, or an EU/EEA certification ANSSI recognises as equivalent: SREN law Art. 31 (2024), décret n° 2026-272 (14 April 2026), arrêté of 12 August 2026; the "cloud au centre" doctrine (2021, updated 2023) extends the expectation to other State IS — exemption regime in those texts
Diffusion Restreinte (DR)Systems accredited per II 901; specific hosting constraints — verify
Classified (Secret / Très Secret, IGI 1300)Dedicated accredited infrastructure; public cloud generally excluded
Banking critical or important functionsAllowed with DORA third-party requirements, outsourcing rules (EBA guidelines) and supervisor notification where required — verify

SecNumCloud: ANSSI qualification of cloud service providers against the référentiel d'exigences v3.2 (8 March 2022; the version approved by the arrêté of 12 August 2026). Beyond technical security requirements, its §19.6 protects against non-EU extraterritorial law: provider established in the EU, non-EU shareholders capped at 24 % individually and 39 % collectively, data stored and the service administered and supervised from the EU. "Sovereign" marketing labels are not qualification: ask for the qualification certificate and its exact service scope. Extraterritorial exposure: assess non-EU legal access risk (e.g., US CLOUD Act, FISA s.702) against data class; encryption with customer-held keys (HSM/BYOK/HYOK) reduces but does not always remove it. Reversibility clauses (contract + architecture): exit plan and assistance period; data and metadata export in documented open formats at no prohibitive cost; deletion certificate; documentation and configuration export; continuity of service during exit; notice periods; sub-processor list; audit rights. EU Data Act (Regulation (EU) 2023/2854, Chapter VI) applies since 12 September 2025: providers must support switching and may bill only their actual switching/egress costs until 12 January 2027, none after that date. Architecture enablers: portable workloads (containers, IaC), standard databases or documented migration paths, avoid deep coupling to proprietary PaaS for core functions unless an ADR accepts it, periodic exit test (at least export + restore elsewhere). Cloud migration options (6R/7R): retire, retain, rehost, relocate, replatform, refactor, repurchase — decide per application from portfolio assessment (see urbanisation reference).

7. Technical debt management

  • Types: architecture (wrong pattern, coupling), obsolescence (end-of-support versions), code, data (quality, duplicates), operational (manual steps), documentation, security (deferred fixes).
  • Register: item, type, location, cause (deliberate via ADR / accidental), interest (cost or risk per period), principal (cost to fix), owner, deadline, link to ADR/dispensation.
  • Obsolescence calendar: vendor end-of-support dates for OS, DB, middleware, runtimes, ERP versions, IBM i releases; flag anything ending within 24 months.
  • Funding: reserve explicit capacity (commonly cited 15-20 % of delivery capacity — organisation-specific); debt from a project's dispensation is costed in that project.
  • Debt affecting critical functions is an ICT risk: report it in the risk framework, not only in IT backlogs.

8. DORA ICT risk from an architecture perspective

Regulation (EU) 2022/2554, applicable since 17 January 2025, completed by RTS/ITS. Article numbers below are indicative: confirm them on EUR-Lex before citing them in a deliverable.

DORA topicArchitecture contribution
ICT risk management framework (Art. 6)Architecture principles, standards and target architecture as documented controls
Identification (Art. 8)Inventory of ICT assets and information assets; mapping of business functions, especially critical or important functions (CIF), to applications, infrastructure, flows and third parties; dependency maps kept up to date
Protection & prevention (Art. 9)Segmentation, IAM, encryption, change management, secure configuration
Detection (Art. 10)Observability and anomaly detection designed in
Response & recovery (Art. 11)Continuity design per CIF, RTO/RPO, degraded modes
Backup, restoration (Art. 12)Backup segregation, restore tests, secondary processing capability where required
Incident classification and reporting (Art. 17-19)Data needed for classification (clients affected, duration, data loss) must be observable
Resilience testing (Art. 24-27)Testable architecture; TLPT for designated entities (Art. 26)
ICT third-party risk (Art. 28-30)Register of information; concentration risk; exit strategies for services supporting CIF; key contractual provisions

Rule: any architecture decision touching a CIF must state its DORA impact (dependencies, third party, exit, RTO/RPO) in the ADR. For non-financial essential/important entities, the equivalent driver is NIS2 (Directive (EU) 2022/2555; French transposition status — verify); for vital operators, sector LPM/OIV rules — verify.

9. Cross-expert coherence review checklist

Run before G1/G2 and on any consolidated deliverable produced by several experts.

  • Same NFR figures everywhere (DAT, infra sizing, DB replication, scheduling windows, contracts/SLAs).
  • RPO consistent with DB replication mode (dba-expert), backups (infra-expert), and batch restart points (scheduling-expert).
  • RTO includes failover, data repair, batch catch-up, partner resynchronisation.
  • Flow matrix = interface contracts = firewall requests = middleware configuration (integration-expert, infra-expert).
  • Each data entity has one master; ERP vs legacy vs MDM ownership explicit (erp-expert, ibmi-expert).
  • Character sets and formats aligned end-to-end (CCSID/EBCDIC ↔ UTF-8, dates, decimals, time zones).
  • Identity and secrets: service accounts, certificates (expiry owners), vault usage, rotation (devsecops-expert).
  • Batch dependencies cross-system and cut-off times consistent with online availability and partner calendars (scheduling-expert).
  • Capacity: volumes and growth used by infra, DB and middleware sizing are the same numbers.
  • Monitoring: every flow and job has an alert, an owner and a run-book entry (DEX).
  • Environment parity: pre-production gaps listed with residual risk.
  • Data classification and hosting location consistent with §6; non-prod data anonymised.
  • Licences: models compatible with the topology (cores, HA/DR nodes, cloud).
  • Reversibility and exit plan exist for each external service; DORA register impact stated for CIF.
  • Every deviation from principles has an ADR or a signed dispensation with expiry.
  • Decommissioning of replaced components planned (data archive, interfaces, licences).

10. Review output format (default)

Findings table: # | area | finding | severity (Critical/High/Medium/Low) | principle/NFR breached | recommendation | owner. Then decision: compliant / with reservations / dispensation / non-compliant, followed by the team-standard Risks, Handoffs, Open questions, Assumptions sections.

Sources

Individual skills in this repo

This repo contains 4 individual skills — each has its own dedicated page.

Crapule421/tiger-team

Write an Architecture Decision Record (ADR) or a target-architecture note from a question or context - options, knock-out and weighted criteria, trade-offs, decision, consequences, reversibility and compliance impacts (security/accreditation, GDPR, DORA, sovereignty). Use when a structural technical choice must be made, documented or reviewed, or when a perimeter needs a documented target state.

Crapule421/tiger-team

Review a batch chain - scheduler export (VTOM, Dollar Universe, Control-M) or a textual description - for dependencies, restartability, windows and critical path, resources, calendars, monitoring and operational risks, with prioritized findings and a restartability matrix. Use before go-live, before a scheduler migration, after a batch incident, or when a chain keeps missing its window.

Crapule421/tiger-team

DBA knowledge base for Oracle, PostgreSQL, SQL Server, Db2 LUW and MySQL/MariaDB - performance diagnosis, locking, HA/DR, backup/PITR, security/GDPR, Oracle-to-PostgreSQL migration pitfalls, release script best practices and a review checklist. Load it when diagnosing a database issue, designing HA/backup, preparing or reviewing a SQL release, or planning a migration.

Crapule421/tiger-team

Fill-in Markdown templates for integration project deliverables (DAT architecture document, DEX operations manual, production release sheet, interface contract, flow matrix, RACI, RAID log, UAT sign-off) and the writing rules that apply to all of them. Load when an agent must produce, update or review a project deliverable.

Related Skills