Skip to content

PHASE-NEXT.md

Latest commit

 

History

History
101 lines (79 loc) · 4.25 KB

PHASE-NEXT.md

File metadata and controls

101 lines (79 loc) · 4.25 KB
Raw

PHASE-NEXT — What's Actually Queued

Operator-facing snapshot of where the implementation is, what's next, and what's blocked. Pulls from the canonical living roadmap at inkmesh/Agent Covenant Protocol_ACP/ACP_Roadmap.md (v0.4.3, last spec update 2026-05-06) — does not fork it.

Updated: 2026-05-09. If this page disagrees with the roadmap doc, the roadmap doc wins; this page is a surface for operators who don't want to read 605 lines.

TL;DR

Phase Status Action
1, 2, 2.5, 3.0, 3.B, 3.A ✅ shipped none
4.1, 4.3, 4.5, 4.6 ✅ shipped none
4.2 — Public launch + observation 🟡 ready, 1-week milestone Do — see PHASE-4.2-LAUNCH.md
7.A — Escrow + Auto-Settlement 🟠 spec drafting (ACR-500 v0.1) Decide 10 working-group questions — see PHASE-7A-DECISIONS.md
7.B / 7.C / 7.D 🔒 gated on 7.A landing + first real escrow tx wait
5 — Cross-Covenant reputation 🔒 gated on 7.A real-transaction data wait — by design
6 — Genesis migration 🔒 gated on first mature OSS adopter requesting onboarding wait
Deferred items (KMS adapters, Redis budget, similarity_threshold) 🔒 gated on demand wait

What "next" means in practice

Two tracks run in parallel right now. They're independent — neither blocks the other.

Track A — Phase 4.2 launch (operator-driven)

Phase 4.2 is not code. It's a one-week observation window: open the server up, watch who shows up, see what the first inbound issues look like, decide whether to pivot scope based on actual demand vs. hypothetical demand.

The roadmap defines this as:

1. 誰部署了？        → GitHub star profile 分類：個人 / 組織
2. 第一個 issue 是什麼？ → 部署 / feature / 合規
3. 三週後決策分流：
   - 多數個人 OSS → 保持本 roadmap 路線
   - 多數企業    → KMS adapter 從延後項提前到 Phase 4
   - 沒 inbound  → 問題在 distribution，roadmap 不動、focus 切 outreach

Concrete steps: see PHASE-4.2-LAUNCH.md.

Track B — Phase 7.A escrow (working-group-driven)

Phase 7.A is the protocol's biggest jump: from "tracks contribution" to "holds and releases real money." That requires ten implementation-level decisions before any code lands. They sit in the spec repo as ACR-500_Decisions_v0.1.md, surfaced for review in PHASE-7A-DECISIONS.md.

Ratifying ≥80% of these decisions unblocks coding. Ratifying <50% means we're still in design space; don't start an implementation that will need to be reworked.

What's NOT next (deliberately)

The roadmap is clear about milestones that wait on real signal:

  • Phase 5 (cross-Covenant reputation) waits on Phase 7.A real transaction data. Reputation built without real money behind it is noise — the roadmap calls this out explicitly. Skip ahead at your peril.
  • Phase 6 (Genesis migration) waits on a specific trigger: a mature open-source project (OpenClaw is the named candidate) asking to on-board their git history. Building this without an adopter ready is a museum piece.
  • KMS / Vault adapters wait on a paying enterprise asking. The KeyProvider interface is already in place (see docs/key-provider.md) — the contract is stable; what's missing is concrete adapters, and we won't build those on speculation.
  • Redis budget counter waits on observed contention in the SQLite atomic UPDATE path. Today there is none.
  • similarity_threshold (ACR-20 Part 4 Layer 3) waits on (a) embedding provider decision and (b) duplicate-content emerging as a measured problem.

Working-agreement on this doc

  • Update on shipped phases. When a phase moves from 🟡 ready / 🟠 spec drafting to ✅ shipped, mark it here and link to the README's status table.
  • Don't add new phases here. New phases are added to the canonical roadmap doc first; this page reflects them after.
  • Don't restate the roadmap's reasoning. Link, don't quote.
  • One page rule. If this gets longer than this version, it's drifting into roadmap-mirror territory — refactor.