Skip to content

Sprint Plan: Finnest — Phase 0 (Foundation)

Date: 2026-04-17 Scrum Master: Gautham Chellappa Project Level: 4 Phase: Phase 0 — Foundation (Weeks 1–4) Total Stories: 20 (STORY-F-021 added 2026-04-18 per Sprint 3 gate check §TD1) Total Points: 52 Planned Sprints: 4 (1 week each)

Workflow context: Generated via BMAD /sprint-planning. Scout + Verify Go-Live (Weeks 5–12) will be re-planned at the start of Week 3 of Phase 0, once execution signal informs the 8-sprint window.

Executive Summary

Phase 0 delivers an empty but deployable Finnest platform — login works, org setup works, agent chat responds, Cmd+K navigates, and the application runs on the existing AgenticAI-app AWS infrastructure (per ADR-014-F co-deploy). Four 1-week sprints totalling 50 points across 19 stories.

Sprint 1 lays the umbrella + local dev baseline. Sprint 2 is the critical decision gate — it validates external MySQL (V2Repo) and deploys to integration. Sprint 3 builds the agent orchestrator + MCP infrastructure + compliance seed. Sprint 4 wires event store immutability, agent chat UI, Cmd+K, and extends the deploy to staging and production hosts.

Key Metrics:

  • Total Stories: 20
  • Total Points: 52
  • Sprints: 4
  • Team: 1 developer (Gautham) + AI assistance
  • Sprint cadence: 1 week, 15-point capacity, ~13-point target velocity
  • Target Completion: 2026-05-15 (4 weeks from 2026-04-17)

Critical milestone: End of Sprint 2 = Week 2 decision gate. If V2Repo + Kamal deploy work, commit to Elixir Direct for Scout + Verify go-live. If blockers emerge, fall back to shipping Laravel as Plan B safety net.

Inputs & References

Sprint plan synthesised from:

Stories are infrastructure / scaffolding — no PRD needed for Phase 0. Scout + Verify PRD covers Weeks 5–12.

Team Capacity

Parameter Value
Team size 1 (Gautham) + AI assistance
Sprint length 1 week (5 working days)
Productive hours/day ~6
Capacity per sprint 30 hours
Point calibration 1 pt ≈ 2–3 hours for Senior + AI (Elixir learning curve absorbs some velocity)
Target capacity 15 points per sprint
Target velocity ~13 points per sprint (buffer for unknowns, Elixir learning)

Capacity model rationale: AgenticAI-app sprint status yaml records 15-point capacity, ~12.5-point rolling velocity. Finnest starts below that baseline due to Elixir learning curve (~30% velocity hit per ADR-0011 tipping-point analysis); recovers to baseline by Sprint 3 as patterns become idiomatic.

Story Inventory

All stories follow the same shape (per BMAD template). Full story files in stories/STORY-F-*.md.

ID Title Epic / Theme Points Priority Sprint
STORY-F-001 Elixir umbrella scaffold (21 OTP app skeletons) Foundation 3 Must 1
STORY-F-002 Boundary library config + CI enforcement (AR-07) Foundation 2 Must 1
STORY-F-003 finnest_core: Repo, correlation ID, tenant primitives, supervisor Foundation 3 Must 1
STORY-F-004 finnest_web: Phoenix endpoint, /health, /ready, DaisyUI asset pipeline Foundation 2 Must 1
STORY-F-005 Bitwarden finnest project + GitHub Actions ci-finnest.yml Infrastructure 2 Must 1
STORY-F-006 phx.gen.auth scaffold + Argon2 + TOTP MFA Auth 3 Must 2
STORY-F-007 Organisation + Office + User schemas + architecture tests Core Data 3 Must 2
STORY-F-008 Tenant enforcement: Finnest.Repo.prepare_query + Tenant plug Multi-Tenancy 2 Must 2
STORY-F-009 V2Repo MyXQL connection to admin_central + admin_atslive (read-only) Migration Foundation 3 Must 2
STORY-F-011 Kamal deploy.finnest.integration.yml + first integration deploy Infrastructure 2 Must 3
STORY-F-012 finnest_agents: Orchestrator + AgentSupervisor + BudgetGuard Agent Infra 3 Must 3
STORY-F-013 AiProvider behaviour + AnthropicDirect + MockProvider adapters Agent Infra 3 Must 3
STORY-F-014 MCP Tool behaviour + ToolRegistry + sample tool Agent Infra 3 Must 3
STORY-F-015 finnest_compliance schemas + credential_types seed (~100 types) Compliance Seed 3 Must 3
STORY-F-016 Event store table + immutability trigger + hash chain Core Data 3 Must 4
STORY-F-017 Agent Chat LiveView + Phoenix Channels streaming Agent UI 3 Must 4
STORY-F-018 Command Bar (Cmd+K) overlay on authenticated pages UX 2 Must 4
STORY-F-019 Industry profiles seed (labour_hire + construction) + ETS cache Compliance Seed 2 Must 4
STORY-F-020 Staging + Production host Postgres + Finnest deploy + smoke test Infrastructure 3 Must 4
STORY-F-021 agents schema + real persistence (sessions, messages, budget, audit) Agent Infra 2 Must 4

Point distribution: 2-pt × 8 stories, 3-pt × 12 stories, no 5+ (all within sprint-fit range).

Sprint Allocation

Sprint 1 — Weeks 1 (2026-04-17 → 2026-04-24)

Goal: Elixir umbrella running locally with Postgres 17 and green CI; health endpoints reachable in a local container.

Story Title Points
STORY-F-001 Elixir umbrella scaffold 3
STORY-F-002 Boundary library + CI enforcement 2
STORY-F-003 finnest_core foundations 3
STORY-F-004 finnest_web Phoenix endpoint + health 2
STORY-F-005 Bitwarden + GitHub Actions ci-finnest.yml 2

Total: 12 / 15 points (80% utilisation; 3-point buffer)

Sprint 1 deliverable: mix test green. Local docker compose up serves /health on port 4000. CI green on push.

Risks: Elixir toolchain setup unfamiliarity (first sprint). Mitigation: AI pair-programs the scaffold; reference AgenticAI-poc-elixir repo.

Sprint 2 — Week 2 (2026-04-24 → 2026-05-01) ⚠️ DECISION GATE

Goal: Authentication works end-to-end; V2Repo reads admin_central; first Finnest deploy to integration host running alongside Laravel.

Story Title Points
STORY-F-006 phx.gen.auth + Argon2 + TOTP MFA 3
STORY-F-007 Organisation + Office + User schemas 3
STORY-F-008 Tenant enforcement prepare_query + plug 2
STORY-F-009 V2Repo MyXQL read-only connection 3

Total: 11 / 15 points (73% utilisation)

Sprint 2 deliverable: End of sprint = Week 2 decision gate. - ✅ Login/logout works with MFA optional - ✅ Org admin can create orgs + users + offices - ✅ V2Repo reads 1 candidate from admin_atslive (proves MyXQL path) - ✅ Tenant isolation architecture test passes (query without context raises)

Gate decision criteria (2026-05-01): - Proceed to Sprint 3 on Elixir track if ALL of above pass - Pivot to Laravel Plan B if MySQL connection or deploy pipeline blocks

Risks: V2Repo MyXQL edge cases; Ecto multi-repo wiring unfamiliar. Mitigation: document decision gate criteria ahead of Sprint 2 start.

Sprint 3 — Week 3 (2026-05-01 → 2026-05-08)

Goal: Agent orchestrator + MCP registry + credential registry seeded; first deploy pipeline runs on every merge.

Story Title Points
STORY-F-011 Kamal deploy.finnest.integration.yml + integration deploy 2
STORY-F-012 finnest_agents orchestrator + supervisor + budget guard 3
STORY-F-013 AiProvider behaviour + AnthropicDirect + Mock 3
STORY-F-014 MCP Tool behaviour + ToolRegistry + sample tool 3
STORY-F-015 finnest_compliance + 100 credential types seeded 3

Total: 14 / 15 points (93% utilisation)

Sprint 3 deliverable: Agent orchestrator classifies a test intent and calls a sample MCP tool end-to-end. Credential registry queryable. Scout + Verify PRD re-planning session scheduled for Sprint 3 end.

Risks: MCP protocol design decisions surface during Sprint 3 (JSON-RPC fallback path, tool registration ergonomics). Mitigation: keep same-node function-call transport (Part 6 decision); defer JSON-RPC transport work.

Sprint 4 — Week 4 (2026-05-08 → 2026-05-15)

Goal: Deployable to all three environments (integration, staging, production-finnest) with agent chat UI + Cmd+K reachable. Phase 0 deliverable demonstrable.

Story Title Points
STORY-F-016 Event store + immutability trigger + hash chain 3
STORY-F-017 Agent Chat LiveView + Phoenix Channels streaming 3
STORY-F-018 Command Bar (Cmd+K) overlay 2
STORY-F-019 Industry profiles seed + ETS cache 2
STORY-F-020 Staging + Production host Postgres + deploy + smoke test 3
STORY-F-021 agents schema + real persistence (swaps F-012/F-013/F-014 stubs) 2

Total: 15 / 15 points (100% utilisation — added 2026-04-18 per Sprint 3 gate check §TD1)

Sprint 4 deliverable: Phase 0 acceptance criteria: - ✅ Login → org setup → agent chat responds → Cmd+K navigates - ✅ Event store rejects UPDATE and DELETE at DB trigger level - ✅ Hash chain verification script runs on event log and passes - ✅ All three envs respond on their respective URLs - ✅ 2 industry profiles loaded (labour_hire + construction) - ✅ Smoke test journey passes on production-finnest

Risks: Staging/prod host resource contention with Laravel (ADR-014-F Option A). Mitigation: monitor RAM during Sprint 4 deploys; upgrade instance if sustained >85% RAM.

Epic / Theme Traceability

Theme Stories Points Sprint(s)
Foundation (umbrella + web + core) F-001, F-002, F-003, F-004 10 1
Infrastructure (secrets, CI, deploy, hosts) F-005, F-011, F-020 7 1, 3, 4
Auth (phx.gen.auth + MFA) F-006 3 2
Core Data (schemas, event store) F-007, F-016 6 2, 4
Multi-Tenancy F-008 2 2
Migration Foundation (V2Repo) F-009 3 2
Agent Infra (orchestrator, AiProvider, MCP, persistence) F-012, F-013, F-014, F-021 11 3, 4
Compliance Seed F-015, F-019 5 3, 4
Agent UI + UX (chat, Cmd+K) F-017, F-018 5 4
Total 20 stories 52 points 4 sprints

Architecture Decision Coverage

Each Phase 0 story implements or directly enables these ADRs:

ADR Coverage in Phase 0
ADR-001-F Elixir/Phoenix stack STORY-F-001 (umbrella), F-004 (Phoenix endpoint)
ADR-002-F Supervised Modular Monolith STORY-F-001 (21 OTP apps), F-002 (Boundary enforcement)
ADR-003-F Three-tier agents STORY-F-012 (orchestrator + supervisor)
ADR-004-F MCP at every domain boundary STORY-F-014 (MCP behaviour + ToolRegistry)
ADR-005-F Event-driven cross-domain STORY-F-016 (event store)
ADR-006-F Hexagonal ports STORY-F-013 (AiProvider port with 2 adapters)
ADR-010-F Strangler Fig migration STORY-F-009 (V2Repo read-only)
ADR-013-F 42 Commandments Every story (test-first, gold-standard files, etc.)
ADR-014-F Infrastructure reuse STORY-F-005, F-011, F-020 (Bitwarden, Kamal, host provisioning)

Three ADRs are built upon but not directly implemented in Phase 0 (deferred to Scout + Verify window):

  • ADR-007-F Three-layer IRAP — Phase 3 work
  • ADR-008-F Flutter mobile — Phase 2 work
  • ADR-011-F Compliance auto-blocking — uses compliance registry from F-015 but the gated caller sites land in Scout/Verify/roster/timekeep sprints
  • ADR-012-F Adopt all features — scoping decision, not implementation

Risks and Mitigation

High:

  • Week 2 decision gate risk (Sprint 2 end) — if MySQL V2Repo or Kamal deploy fail, entire Elixir Direct plan is at risk. Mitigation: Laravel Plan B ship stays an option through Sprint 2; explicit decision criteria documented at Sprint 2 kickoff.
  • Elixir learning curve — Sprint 1 velocity likely below target. Mitigation: AI-pair programming; leverage AgenticAI-poc-elixir repo as pattern reference; ~30% velocity buffer built into capacity.

Medium:

  • ADR-014-F co-deploy resource contention on t3.medium hosts. Mitigation: Monitor RAM; upgrade to t3.large before Phase 1 end if needed.
  • IRAP pre-assessment scheduling — should happen in Phase 0 per brainstorm-10 but is a separate thread. Mitigation: Gautham schedules IRAP assessor meeting independently of sprint flow; not blocking Phase 0.

Dependencies

  • External: IRAP assessor availability (non-blocking for Phase 0); Anthropic API key; AWS account access for all 4 envs.
  • Internal: Bitwarden Secrets Manager admin access for finnest project creation; GitHub Actions runner availability on CI bastion; Route 53 access for subdomain records.
  • Cross-repo: None — Finnest code repo (MMGS-Softnet-Pty-Ltd/finnest) not created yet; code lives in a working copy during Phase 0, published to the code repo when Sprint 4 deploys land.

Definition of Done

A Phase 0 story is complete when:

  • Code implemented in the Finnest Elixir working copy
  • Unit + integration tests written (ExUnit) — target ≥80% coverage per CQ-02/CQ-03
  • mix format --check-formatted passes (CQ-01)
  • mix credo --strict passes (CQ-20)
  • mix dialyzer zero warnings (CQ-09)
  • mix sobelow + mix deps.audit green (SE-11)
  • mix boundary passes (AR-08) — no cross-domain violations
  • Architecture tests (tenant isolation, event immutability) still pass (once applicable)
  • Relevant acceptance criteria in this story file all ticked
  • Story status updated in sprint-status.yaml
  • Commit references story ID (STORY-F-NNN: short message)

Worktree Quickstart

JIT — copy-paste these when you start each story, not now. Each command creates a feature-branch worktree under ~/Documents/GitHub/finnest-worktrees/ and cds into it. From there: claude → /first-in → /dev-story.

All Phase 0 stories are backend (no mobile stories until Phase 2).

# blocked by ... annotations show within-sprint blockers only — cross-sprint blockers are implied satisfied by sprint ordering. For the live wave-by-wave view, run /moons-spawn.

Sprint 1 — Foundation

gwtnew-backend F-001-elixir-umbrella-scaffold
gwtnew-backend F-002-boundary-ci-enforcement       # blocked by F-001
gwtnew-backend F-003-finnest-core-foundations      # blocked by F-001, F-002
gwtnew-backend F-004-finnest-web-endpoint          # blocked by F-003
gwtnew-backend F-005-bitwarden-gh-actions-ci       # blocked by F-001, F-002

Sprint 2 — Auth + V2Repo + Deploy (Decision Gate)

gwtnew-backend F-006-phx-gen-auth-argon2-totp
gwtnew-backend F-007-org-office-user-schemas       # blocked by F-006
gwtnew-backend F-008-tenant-enforcement            # blocked by F-007
gwtnew-backend F-009-v2repo-myxql-connection

Sprint 3 — Agent Infra + Integration Deploy

gwtnew-backend F-011-kamal-integration-deploy
gwtnew-backend F-012-finnest-agents-orchestrator
gwtnew-backend F-013-ai-provider-adapters          # blocked by F-012
gwtnew-backend F-014-mcp-tool-registry             # blocked by F-012
gwtnew-backend F-015-finnest-compliance-schemas-seed

Sprint 4 — Event Store + Agent UI + Full Deploy

gwtnew-backend F-016-event-store-immutability-hash-chain
gwtnew-backend F-021-agents-schema-persistence          # blocked by F-012 (Sprint 3)
gwtnew-backend F-017-agent-chat-liveview                # blocked by F-016, F-021
gwtnew-backend F-018-command-bar-cmdk-overlay           # blocked by F-017
gwtnew-backend F-019-industry-profiles-seed-ets-cache
gwtnew-backend F-020-staging-prod-deploy-smoke-test     # blocked by F-016, F-017, F-018, F-019

Next Steps

Immediate: Start Sprint 1.

Primary actions:

  1. Start implementing STORY-F-001 (umbrella scaffold) — use /dev-story STORY-F-001 or work independently
  2. Sprint 1 retrospective end of Week 1 — record actual velocity in sprint-status.yaml
  3. Schedule IRAP pre-assessment meeting (independent thread, Gautham-owned)
  4. Schedule Scout + Verify PRD re-planning session for 2026-05-08 (Sprint 3 end)

Sprint cadence:

  • Sprint length: 1 week
  • Sprint kickoff: Monday
  • Sprint review/retrospective: Friday
  • Sprint planning for next sprint: Friday (rolling — re-plan against reality)

After Phase 0: Sprint planning session for Scout + Verify Go-Live (Weeks 5–12) — 8 sprints covering the PRD's 10 epics and ~60 stories. Run /sprint-planning again at Week 3 of Phase 0 with Phase 0 execution signal informing capacity and risk.

This plan was created using BMAD Method v6 — Phase 4 (Implementation Planning)