Initial MVP evidence and superseded implementation plan

ARCHIVED: this file preserves original implementation evidence and the earlier signed-entity/xitdb revision proposal. Its pending checklists are not active work. The owner subsequently chose SQLite with explicit change events, direct FIPS attribution and dependency-aware incoming review. The current authoritative plan is implementation-plan.md. No historical checkmark establishes completion of the new design. Relative links below remain for historical context.

Status: initial MVP implemented and locally checked; the owner approved the revised product/architecture below after first deployed review. Revision work is not yet implemented. Successful deployed sync remains unverified. App alias: teams-todo, title Teams Todo, protocol teams-todo.sync.v1. Earlier notes called it todo-teams; use teams-todo consistently for the new scaffold.

Dependencies: core C3 contracts, then FIPS signing/owner-setup/invitation revision, building on the existing F1–F3 baseline. The agreed boundaries and soft-grant policy are in the architecture.

This plan moved into the independent app repository after scaffolding. The core docs retain a short entry point. All checked boxes below describe code/local checks, not authenticated-browser or real-network verification. R0–R7 below are the active execution plan. The original M0–M5 record is retained as historical implementation evidence; its draft-every-edit, unsigned events, automatic publication and manual descriptor/grant setup policies are superseded, not requirements for new work.

First deployed review

See UX and sync investigation for the user's reported setup difficulties, source findings, read-only local state evidence and proposed improvements. The inspected local board has no Teams Todo grants/requests and a saved permission-denial error; no app contact/durable sync is recorded. This is a confirmed local blocker, not proof of a binary transport/storage defect. VPS state and actual post-approval replication remain unverified. The owner subsequently approved node signatures, manual publication, integrated create/join grants, direct invitations, separate board/task views, diagnostics and explicit reset tools. Only the shortcut icon correction has been implemented from this review so far.

Approved product decisions

Execution order and gates

R0 pins the contracts/storage boundary; R1 depends on the core/FIPS signing work; R2 builds the publication boundary on signed data. R3 depends on FIPS owner setup and invitation transport. R4/R5 integrate the new UI and diagnostics, R6 reset lifecycle, then R7 deployed acceptance. Work on presentation must use the confirmed structure below; propose any substantial departure before coding.

This plan authorizes documentation of the direction, not implementation, publishing, commits, pushes, hooks or automated tests. Future implementation follows explicit user instructions and the existing targeted lint/one-off-check rules.

R0 — pin contracts, state model and reset transition

Exit: a concrete signed schema, staged-publication/crash-recovery design and bounded onboarding contracts, with unresolved details recorded before code depends on them.

R1 — independently verifiable node-authored data

Owning files: model.ts, storage.ts, app.ts; FIPS owns the signing key/service.

Exit: stored shared source data is independently node-verifiable, even outside the connection that delivered it. Human/bunker signatures and forwarded histories remain out of scope; direct-origin replication is retained for this increment.

R2 — immediate local edits, explicit publication

Owning files: app.ts, storage.ts, network.ts, model.ts.

Exit: a user can edit freely while peers cannot retrieve those edits until Send updates. Published history remains immutable; corrections are new events.

R3 — guided create/join and direct invitations

Owning files: app.ts, init.ts, network.ts, storage.ts, definition.ts, view.ts.

Exit: paired owners can create and join from the app without copying descriptors, protocol IDs, board IDs or grant JSON. Each owner makes one contextual setup approval.

R4 — boards-only navigation and focused task UI

Exit: the board is understandable without knowing any ID, and task interaction resembles personal Todo while sharing metadata is available without dominating it.

R5 — truthful status, activity and bounded diagnostics

Exit: UI answers saved/sent/received/blocked/next-action; logs explain failures at the correct layer without claiming a transport write is durable app receipt.

R6 — clear, remove and repeatable fresh-start workflows

Exit: repeatable clean setup is available without deleting unrelated data or allowing retained remote histories to repopulate a removed board silently.

R7 — revised acceptance and release handoff

Historical initial MVP record

The following evidence and M0–M5 checklists describe the original implementation, not completion of R0–R7. Where they conflict with the approved revision above, the revised decisions are authoritative.

Implementation evidence

Original product slice (superseded where noted)

Mimic the personal Todo interaction style with only:

No AI task generation, subtask hierarchy, priority duels, dependencies, attachments, comments, notifications to humans, role hierarchy, billing, generic sync framework, automatic peer introductions or routed mesh in this MVP. Paid teams-* apps are a future product direction, not a licensing implementation prerequisite.

The app owns membership, data validation, subscription state, conflict resolution and optional future Git-style review. None are FIPS/capability responsibilities.

M0 — pin storage and validate the replication slice

xitdb-ts is the first candidate, not an unverified promise. Pin a concrete release / commit and inspect its documented/source behavior before designing file-level sync.

Exit: evidence for resumable committed-byte replication and explicit limitations, not a claim that immutable moments automatically provide event sourcing or merge.

M1 — scaffold and local Todo slice

Use the existing helper, inspected at src/plugin-lifecycle/scaffold.ts:219:

createPluginScaffold({
  dmBotRoot: <AppWeaver root>,
  alias: "teams-todo",
  title: "Teams Todo",
  description: "Local-first shared todo boards over FIPS",
  coreApiVersion: <current core range>,
  createDraftCommands: true,
  runGenerator: true,
  configureGitHooks: false
})

The helper creates the independent Git repository, adds the local manifest entry and generates registration. This invocation explicitly skipped hook configuration; no hooks were run. It does not publish the app or approve networking. This invocation is complete; do not rerun over the existing plugin directory. Replace template sample behavior rather than exposing scaffold demo tools as app features.

M2 — board membership and deterministic event model

These are proposed small-MVP policies, not requirements imposed on other apps:

Checklist:

M3 — direct incremental sync over binary messages

Protocol teams-todo.sync.v1; exact topic is the board ID. Every request/response uses an explicit peer FIPS key and the same protocol/topic grant boundary.

App-owned message families:

Message Meaning
sync-state Writer/log identity and current committed replica cursor
sync-start Validated source identity, byte range, transfer ID and size bounds
sync-chunk Transfer ID, expected offset and unchanged source bytes
sync-complete Transfer end; validate and commit the replica
sync-ack Confirm durable committed position, not UI acceptance/merge
sync-error App-protocol failure with bounded diagnostic reason

Use a small app binary header for chunk identity/offset plus raw xitdb bytes; control messages can use text. This is above FIPS framing, not a second transport.

M4 — app-owned subscriptions and near-real-time hints

M5 — three-node acceptance and MVP publication

No automated test suite is added/run unless requested. Narrow in-memory validation and user-assisted deployed scenarios supply initial evidence; authenticated browser automation requires user-confirmed access. Document existing vendor failures separately from new code errors.

Later app work

Git-style incoming change review, shared acceptance/resolution events, selective project roles, retained-history/compaction policy, cross-peer forwarding of writer logs with verifiable original authorship, AI features and paid licensing can follow. Do not equate network forwarding with database-history forwarding.

Progress