@damphuquy/agent-init 1.4.4 → 2.1.0

package/README.md

@@ -1,6 +1,6 @@
 # agent-instructions (@damphuquy/agent-init)
 
-> **AI Coding Agent Scaffolding & Orchestration System (RIPER-5 Framework)**
+> **AI Coding Agent Scaffolding & Orchestration System v2.0 (RIPER-5 Framework)**
 
 <p align="center">
   <b>English</b> • <a href="README.vi.md">Tiếng Việt</a>
@@ -8,7 +8,13 @@
 
 ---
 
-A structured scaffolding framework for AI Coding Agents following the **RIPER-5 Framework** — cleanly separated into three layers: `AGENTS.md` (project configuration), `.agents/` (agent behavioral controls), and `process/` (task execution and verification workflow).
+A structured scaffolding framework for AI Coding Agents following the **RIPER-5 Framework v2.0** — cleanly separated into three layers: `AGENTS.md` (project configuration), `.agents/` (agent behavioral controls), and `process/` (task execution and verification workflow).
+
+Featuring four core **v2.0 Engineering Protocols**:
+* **Subagent Orchestration v2.0:** Parallel vertical slicing with zero write collision and reactive wakeup.
+* **Implementation Standards v2.0:** Strict typing (zero `any`), domain purity, fail-fast boundary validation, and deterministic teardown.
+* **Planning & Slicing Standards v2.0:** INVEST-compliant vertical slices (2–5 per task) with mandatory automated verifiers.
+* **Testing Standards & Harness v2.0:** Deterministic test pyramid with frozen clocks, fixture factories, and strict mocking boundaries.
 
 ```bash
 # Bootstrap immediately into your project (takes 30 seconds)
@@ -23,11 +29,11 @@ Designed for minimal overhead, practical focus, and zero fluff:
 
 ### 1. For Daily Development (Start Here)
 * **[01 · Hands-on Walkthrough from A–Z](docs/en/01-getting-started.md)**: **The core guide to get started in 10–15 minutes**. Step-by-step instructions with **2 practical examples** (Task 1 Strict Track $\rightarrow$ Handoff $\rightarrow$ Task 2 Fast-Track with knowledge inheritance), prompt templates, and quality gate sign-offs.
-* **[02 · Daily Cheat Sheet](docs/en/02-cheat-sheet.md)**: Copy-paste prompt templates for each phase, quality gates summary (Gates G0–G3), and file ownership matrix (Human vs Agent).
+* **[02 · Daily Cheat Sheet](docs/en/02-cheat-sheet.md)**: Copy-paste prompt templates for each phase, quality gates summary (Gates G0–G3), subagent delegation prompts, and v2.0 protocol matrix.
 * **[03 · Advanced Workflows & Production Edge Cases](docs/en/03-advanced-workflows.md)**: Complex real-world scenarios: Epics in `process/features/`, multi-developer concurrency, git rebase conflicts, P0 emergency hotfixes, secret scanning, model switching, complex test harnesses, and task parking (`paused.md`).
 
 ### 2. In-Depth Reference (For Tech Leads & Maintainers)
-* **[3-Layer Architecture & Context Router](docs/en/reference/architecture.md)**: Deep dive into the 3-layer decoupled architecture and Minimum Sufficient Context principles (No Full-Repo Scan).
+* **[3-Layer Architecture & Context Router (v2.0)](docs/en/reference/architecture.md)**: Deep dive into the 3-layer decoupled architecture, the 11-seed catalog, and the v2.0 Engineering Protocols.
 * **[DevOps & Release Automation](docs/en/reference/devops-release.md)**: CI/CD release lifecycle, automated Semantic Versioning, and NPM package publishing.
 
 ---

package/README.vi.md

@@ -1,6 +1,6 @@
 # agent-instructions (@damphuquy/agent-init)
 
-> **Hệ thống điều phối và khởi tạo AI Coding Agent theo chuẩn RIPER-5**
+> **Hệ thống điều phối và khởi tạo AI Coding Agent v2.0 theo chuẩn RIPER-5**
 
 <p align="center">
   <a href="README.md">English</a> • <b>Tiếng Việt</b>
@@ -8,7 +8,13 @@
 
 ---
 
-Hệ thống mẫu (scaffold) điều phối AI Coding Agent theo chuẩn **RIPER-5 Framework** — phân tách thành ba layer rõ ràng: `AGENTS.md` (cấu hình dự án), `.agents/` (quy chuẩn hành vi của Agent), và `process/` (quy trình thực thi task và kiểm thử).
+Hệ thống mẫu (scaffold) điều phối AI Coding Agent theo chuẩn **RIPER-5 Framework v2.0** — phân tách thành ba layer rõ ràng: `AGENTS.md` (cấu hình dự án), `.agents/` (quy chuẩn hành vi của Agent), và `process/` (quy trình thực thi task và kiểm thử).
+
+Tích hợp 4 **Giao thức Kỹ thuật v2.0** cốt lõi:
+* **Điều phối Subagent v2.0:** Chia lát cắt song song, không xung đột ghi (zero write collision) và đánh thức phản ứng.
+* **Quy chuẩn Hiện thực Mã nguồn v2.0:** Kiểm soát kiểu nghiêm ngặt (không `any`), domain thuần khiết, fail-fast tại ranh giới và dọn dẹp tài nguyên xác định.
+* **Quy chuẩn Lập kế hoạch & Cắt lát v2.0:** Lát cắt dọc theo chuẩn INVEST (2–5 slice/task) với lệnh kiểm chứng tự động bắt buộc.
+* **Quy chuẩn Kiểm thử & Test Harness v2.0:** Kim tự tháp kiểm thử xác định với đóng băng thời gian (frozen clocks), fixture factories và mock có kiểm soát.
 
 ```bash
 # Khởi tạo ngay vào project của bạn (chỉ mất 30 giây)
@@ -23,11 +29,11 @@ Tài liệu được thiết kế tối giản, tập trung vào thực hành, k
 
 ### 1. Dành cho công việc hằng ngày (Bắt đầu tại đây)
 * 📘 **[01 · Hướng dẫn Thực chiến từ A–Z](docs/vi/01-getting-started.md)**: **Tài liệu cốt lõi để bắt đầu trong 10–15 phút**. Hướng dẫn từng bước với **2 ví dụ thực tế liên hoàn** (Task 1 Strict Track $\rightarrow$ Bàn giao $\rightarrow$ Task 2 Fast-Track kế thừa tri thức), prompt mẫu và cách duyệt qua các cổng.
-* ⚡ **[02 · Sổ tay Tra cứu Nhanh (Cheat Sheet)](docs/vi/02-cheat-sheet.md)**: Bảng copy-paste prompt mẫu cho từng giai đoạn, bảng 4 cổng kiểm soát (Gates G0–G3), và ma trận quyền hạn file (Human vs Agent).
+* ⚡ **[02 · Sổ tay Tra cứu Nhanh (Cheat Sheet)](docs/vi/02-cheat-sheet.md)**: Bảng copy-paste prompt mẫu cho từng giai đoạn, bảng 4 cổng kiểm soát (Gates G0–G3), prompt ủy quyền subagent và ma trận giao thức v2.0.
 * 🚀 **[03 · Quy trình Nâng cao & Tình huống Thực chiến](docs/vi/03-advanced-workflows.md)**: Các kịch bản production phức tạp: Epic trong `process/features/`, đa lập trình viên đồng thời, xử lý xung đột rebase, hotfix P0 khẩn cấp, quét secret bảo mật, đổi model chat, và đóng băng task dài hạn (`paused.md`).
 
 ### 2. Tài liệu Chuyên sâu & Tham khảo (Dành cho Lead / Maintainer)
-* 🏛️ **[Kiến trúc 3 Tầng & Cơ chế Context Router](docs/vi/reference/architecture.md)**: Giải thích sâu về thiết kế tách rời 3 layer và cơ chế nạp ngữ cảnh tối thiểu (No Full-Repo Scan).
+* 🏛️ **[Kiến trúc 3 Tầng & Cơ chế Context Router (v2.0)](docs/vi/reference/architecture.md)**: Giải thích sâu về thiết kế tách rời 3 layer, danh mục 11 seed template và 4 Giao thức Kỹ thuật v2.0.
 * 🚢 **[Cẩm nang DevOps & Tự động hóa Phát hành](docs/vi/reference/devops-release.md)**: Quy trình CI/CD tự động tính Semantic Versioning và publish package lên NPM.
 
 ---

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@damphuquy/agent-init",
-  "version": "1.4.4",
+  "version": "2.1.0",
   "description": "Scaffolding CLI to bootstrap RIPER-5 Coding Agents & Operational Workspace",
   "type": "module",
   "main": "./dist/index.js",

package/templates/en/.agents/behavior.md

@@ -25,9 +25,41 @@ first line. The declared mode must match the current phase recorded in `task.md`
 Omit the mode declaration only for pure conversational exchanges that do not
 advance a task (e.g., answering a factual question, clarifying scope).
 
+In continuous autonomous execution (DELEGATED / Fast-Track), the agent opens with the starting phase's mode. When transitioning across phases within a single turn, emit an explicit transition marker:
+`>>> [PHASE TRANSITION: <OLD_PHASE> -> <NEW_PHASE>]` and proceed under the new mode immediately without pausing.
+
+---
+
+## 2. Working Modes & Phase Transition Protocol
+
+<working_modes_protocol>
+  Execution posture is determined by `<working_mode>` in `task.md` or user prompt instruction:
+
+  ### PAIR Mode (Default — Step-by-Step Collaboration):
+  - The agent works on one phase at a time.
+  - Upon completing a phase, it updates the corresponding artifact and **HALTS** to let the human engineer inspect, discuss, and sign off the Gate (G1, G2, G3).
+  - Waits for user prompt (e.g., "Start next phase") before updating `<current_phase>` in `task.md` and continuing.
+
+  ### DELEGATED Mode (Autonomous / Fast-Track / Skip Permissions):
+  - **Activation:** `<working_mode>DELEGATED</working_mode>` in `task.md` OR explicit prompt directive ("fast-track", "skip permissions", "run automatically", "auto-advance", "autonomous").
+  - **Core Rule:** **DO NOT HALT AFTER EACH PHASE TO WAIT FOR USER PROMPT "NEXT".**
+  - **Continuous Transition Workflow:**
+    1. When the current phase meets its exit criteria/checklist, the agent checks `- [x]`.
+    2. Populates auto-approval into the artifact: `approved_by: [AUTO: DELEGATED]` with timestamp and technical rationale (in INNOVATE: adopts the optimal Recommendation; in PLAN: locks the scope contract).
+    3. Immediately updates `<current_phase>` in `task.md` to the next phase (`RESEARCH` → `INNOVATE` → `PLAN` → `EXECUTE` → `REVIEW`).
+    4. Instantiates the next phase seed artifact (Copy-On-Demand) and **CONTINUES EXECUTION IMMEDIATELY** within the same session/turn.
+  - **Sole Stop Conditions in DELEGATED:**
+    - Task is 100% COMPLETE (Gate 3 PASS, housekeeping cleaned, `handoff.md` generated, moved to `completed/`).
+    - OR a true Escalation Trigger is tripped (retry budget exhausted after 3 attempts, destructive command, or unresolvable invariant conflict).
+
+  ### MANUAL & DIAGNOSE-ONLY Modes:
+  - `MANUAL`: Human leads command-by-command; agent provides scoped assistance.
+  - `DIAGNOSE-ONLY`: Runs Research & Review for root-cause audit without mutating source code.
+</working_modes_protocol>
+
 ---
 
-## 2. Session Startup Protocol
+## 3. Session Startup Protocol
 
 Before continuing any in-progress task, reload persistent state in this order:
 
@@ -45,7 +77,7 @@ file-based artifacts listed above.
 
 ---
 
-## 3. Context Navigation Rules
+## 4. Context Navigation Rules
 
 <context_rules>
   <rule id="minimum_context">

package/templates/en/.agents/guardrails.md

@@ -39,10 +39,11 @@
      not declared in the approved plan.
   3. Required change touches a file outside the scope defined in `plan.md`.
   4. A business or policy decision is needed that is not in `<approved_decisions>`.
+     (Exception: Under DELEGATED / Fast-Track mode, the agent is authorized to select the recommended technical option evaluated in `decision.md` without triggering escalation, provided it preserves `<invariants>` and `<out_of_scope>`).
   5. Scope expansion is needed beyond `<out_of_scope>` in `task.md`.
 
   Exception: Do NOT halt if the change was explicitly authorized by the user
-  in the prompt/spec, or if it is a mandatory accompanying test or import update.
+  in the prompt/spec (under DELEGATED, fast-track, or skip permissions), or if it is a mandatory accompanying test or import update.
 </escalation_triggers>
 
 ---

package/templates/en/AGENTS.md

@@ -69,6 +69,12 @@
 
 <riper5_protocol>
 
+  <!-- ─────────────────── WORKING MODES ─────────────────── -->
+  <working_modes>
+    <mode id="PAIR" default="true">Step-by-step collaboration. Halts after each phase for human engineer review and gate sign-off.</mode>
+    <mode id="DELEGATED" alias="fast-track,autonomous,skip-permissions">Autonomous delegation. When task.md has working_mode=DELEGATED or the user specifies fast-track / skip permissions, the agent auto-certifies qualifying gates ([AUTO: DELEGATED]), advances task.md current_phase, and executes continuously without pausing for "next phase" prompts.</mode>
+  </working_modes>
+
   <!-- ─────────────────── PHASE CONSTRAINTS ─────────────────── -->
 
   <phase name="RESEARCH" order="1">
@@ -76,15 +82,16 @@
     <constraint>No implementation decisions.</constraint>
     <constraint>No architectural choices.</constraint>
     <output>Produce/update `research.md`. Classify evidence as Confirmed / Observed / Hypothesized.</output>
-    <gate id="G0">All Research Exit Criteria in `research.md` checked before advancing.</gate>
+    <gate id="G0">All Research Exit Criteria in `research.md` checked before advancing. In DELEGATED / Fast-Track, the agent auto-certifies and advances to INNOVATE immediately.</gate>
   </phase>
 
   <phase name="INNOVATE" order="2">
     <constraint>READ-ONLY. No source-code modifications.</constraint>
-    <constraint>Present 2–3 viable options with trade-off matrix. Do NOT decide unilaterally.</constraint>
-    <constraint>Never decide public API / DB schema / security policy / business policy without human approval.</constraint>
-    <output>Produce `decision.md`. Leave `<engineer_decision>` blank — human fills it.</output>
-    <gate id="G1">Gate 1 in `decision.md` checked and signed by engineer before advancing to Plan.</gate>
+    <constraint>Present 2–3 viable options with trade-off matrix.</constraint>
+    <constraint>PAIR mode: Leave `<engineer_decision>` blank for the human engineer to review and complete.</constraint>
+    <constraint>DELEGATED / Fast-Track mode: Automatically adopt the optimal Recommendation, record rationale with `[AUTO: DELEGATED]`, sign Gate 1, and immediately advance to PLAN.</constraint>
+    <output>Produce `decision.md`.</output>
+    <gate id="G1">Gate 1 in `decision.md` signed by engineer (PAIR) or auto-certified by agent with `[AUTO: DELEGATED]` (DELEGATED/Fast-Track) before advancing to Plan.</gate>
   </phase>
 
   <phase name="PLAN" order="3">
@@ -92,7 +99,7 @@
     <constraint>Every slice must have a defined verifier, expected evidence, and rollback point.</constraint>
     <constraint>Scope contract (allowed / forbidden files) must be explicit.</constraint>
     <output>Produce `plan.md`. Populate Verification Matrix headers.</output>
-    <gate id="G2">Gate 2 in `plan.md` checked and signed by engineer before advancing to Execute.</gate>
+    <gate id="G2">Gate 2 in `plan.md` signed by engineer (PAIR) or auto-certified by agent with `[AUTO: DELEGATED]` (DELEGATED/Fast-Track) before advancing to Execute.</gate>
   </phase>
 
   <phase name="EXECUTE" order="4">
@@ -100,7 +107,7 @@
     <constraint>One slice at a time. Run verifier after each slice. Inspect diff after each slice.</constraint>
     <constraint>No unrelated refactoring. No changes to forbidden files.</constraint>
     <constraint>Update `state.md` after every slice.</constraint>
-    <output>Source code + tests + updated `state.md` with verification evidence.</output>
+    <output>Source code + tests + updated `state.md` with verification evidence. When all slices pass, advance directly to REVIEW.</output>
   </phase>
 
   <phase name="REVIEW" order="5">
@@ -108,7 +115,7 @@
     <constraint>No code fixes during review. Log findings in `review.md` instead.</constraint>
     <constraint>Cover: behavior, architecture, data, security, regression.</constraint>
     <output>Produce `review.md` with findings classified by category/severity/type and Gate 3 checklist.</output>
-    <gate id="G3">Gate 3 in `review.md` checked before handoff. Review decision must be PASS.</gate>
+    <gate id="G3">Gate 3 in `review.md` must PASS before handoff. In DELEGATED mode, the agent performs full audit, creates `handoff.md`, moves task to `completed/`, and reports completion.</gate>
   </phase>
 
   <!-- Behavior rules (mode declaration, retry, escalation, completion gate) →

package/templates/en/process/README.md

@@ -24,6 +24,7 @@ process/
 │   ├── review-template.md.seed       # Review phase artifact (Gate 3)
 │   ├── handoff-template.md.seed      # Final projection (short)
 │   ├── cancellation-template.md.seed # Knowledge-preserving cancellation record
+│   ├── pause-template.md.seed        # Task pause & parking record (temporary blockages)
 │   ├── results-template.tsv.seed     # Benchmark & quantitative metrics tracker
 │   └── program-template.md.seed      # Multi-phase program blueprint
 ├── context/                          # Durable knowledge base & context routers
@@ -45,6 +46,7 @@ process/
 │   │       ├── review.md             # Review phase output
 │   │       ├── handoff.md            # Final projection
 │   │       ├── results.tsv           # (Optional) Benchmark & metric iteration tracking
+│   │       ├── paused.md             # (If paused) Parking record & resumption criteria
 │   │       └── cancelled.md          # (If aborted) Cancellation findings & rollback status
 │   ├── completed/                    # Archived task workspaces
 │   └── backlog/                      # Backlog notes: {note_slug}_NOTE_{dd-mm-yy}.md
@@ -69,16 +71,16 @@ flowchart TD
         Task --> Res["research.md"] --> G0{"Gate G0 (Auto)"}
     end
     subgraph I2["2. INNOVATE (Read-Only)"]
-        G0 --> Dec["decision.md (2-3 Options)"] --> G1{"Gate G1 (Human Approval)"}
+        G0 --> Dec["decision.md (2-3 Options)"] --> G1{"Gate G1 (Engineer or Auto-DELEGATED)"}
     end
     subgraph P3["3. PLAN (Plan-Only)"]
-        G1 --> Plan["plan.md (Vertical Slices)"] --> G2{"Gate G2 (Human Sign-off)"}
+        G1 --> Plan["plan.md (Vertical Slices)"] --> G2{"Gate G2 (Engineer or Auto-DELEGATED)"}
     end
     subgraph E4["4. EXECUTE (Scoped R/W)"]
         G2 --> Code["Implement Slice"] --> Test["Run Verifiers"] --> State["state.md (Memory)"]
     end
     subgraph R5["5. REVIEW (Read-Only)"]
-        State --> Rev["review.md (Audit)"] --> G3{"Gate G3 (Human Approval)"}
+        State --> Rev["review.md (Audit)"] --> G3{"Gate G3 (Engineer or Auto-DELEGATED)"}
     end
     subgraph Done["6. COMPLETE"]
         G3 --> Arch["Move to completed/"] --> Handoff["handoff.md"]
@@ -88,10 +90,10 @@ flowchart TD
 ```text
 task.md          ← (always) Master contract, RIPER phase/gate state, AC, decisions
 research.md      ← (Research) Execution flow, evidence, boundaries
-decision.md      ← (Innovate) Options, trade-offs, Gate 1 — engineer must approve
-plan.md          ← (Plan) Slices, verifiers, scope contract, Gate 2 — engineer must approve
+decision.md      ← (Innovate) Options, trade-offs, Gate 1 (Engineer approval or Auto-DELEGATED)
+plan.md          ← (Plan) Slices, verifiers, scope contract, Gate 2 (Engineer sign-off or Auto-DELEGATED)
 state.md         ← (Execute) Per-slice progress, failure memory, retry budget
-review.md        ← (Review) Findings, verification matrix, Gate 3
+review.md        ← (Review) Findings, verification matrix, Gate 3 (PASS)
 handoff.md       ← (Complete) Short final projection
 ```
 
@@ -107,8 +109,10 @@ handoff.md       ← (Complete) Short final projection
 | Execute | source code + tests + `state.md` |
 | Review | `review.md` |
 | Handoff | `handoff.md` |
+| Pause / Park | `paused.md` (clean freeze & resumption criteria upon temporary block) |
 | Cancellation | `cancelled.md` (knowledge preservation upon abort) |
 | Benchmark / Metrics | `results.tsv` (performance and eval metrics) |
+| Program | `program.md` (multi-phase umbrella epic blueprint) |
 | Gate 1 | `decision.md` approval |
 | Gate 2 | `plan.md` approval |
 | Gate 3 | `review.md` approval |
@@ -126,14 +130,14 @@ handoff.md       ← (Complete) Short final projection
   </phase>
 
   <phase order="2" name="Innovate">
-    READ-ONLY. Present 2–3 options with trade-off matrix. Do not decide unilaterally.
-    Produce `decision.md`. Engineer approves Gate 1.
+    READ-ONLY. Present 2–3 options with trade-off matrix.
+    Produce `decision.md`. Gate 1: Engineer approves (PAIR mode) or Agent auto-certifies optimal recommendation [AUTO: DELEGATED] (DELEGATED/Fast-Track mode).
   </phase>
 
   <phase order="3" name="Plan">
     Plan artifacts only — no source changes.
     Decompose into vertical slices with verifiers, rollback points, and scope contract.
-    Produce `plan.md`. Engineer approves Gate 2.
+    Produce `plan.md`. Gate 2: Engineer approves (PAIR mode) or Agent auto-certifies scope contract [AUTO: DELEGATED] (DELEGATED/Fast-Track mode).
   </phase>
 
   <phase order="4" name="Execute">
@@ -143,7 +147,7 @@ handoff.md       ← (Complete) Short final projection
 
   <phase order="5" name="Review">
     READ-ONLY. May run verification commands. No code fixes.
-    Produce `review.md` with findings. Gate 3 must pass before handoff.
+    Produce `review.md` with findings. Gate 3 must pass before handoff. Under DELEGATED mode, Agent auto-completes housekeeping and creates `handoff.md`.
   </phase>
 </operational_phases>
 

package/templates/en/process/_seeds/_GUIDE.md

@@ -12,7 +12,7 @@
   ### Seeds Architectural Rationale: Understanding "Why" & "How"
   * **Why the `.seed` extension? (Why):** The `.seed` suffix acts as an immutable boundary. It prevents automated AI discovery tools (`find`, `grep`) from confusing archetype templates with active `*.md` task files. This guarantees agents will never accidentally overwrite master blueprints during execution.
   * **Why separate seeds per phase? (Why):** Each phase of RIPER-5 requires a distinct cognitive posture and permission boundary (Research vs Innovate vs Plan vs Execute vs Review). Splitting archetypes prevents context window bloat (Anti-Context Saturation), avoids hallucinations, and anchors each Quality Gate (G1–G3) to an auditable physical artifact.
-  * **Copy-On-Demand Protocol (How):** Never bulk-copy all seeds into a task directory. Start exclusively with `task.md`. Sequentially instantiate subsequent artifacts (`research.md` $\rightarrow$ `decision.md` $\rightarrow$ `plan.md` $\rightarrow$ `state.md` $\rightarrow$ `review.md` $\rightarrow$ `handoff.md`) only as the task progresses into each phase.
+  * **Copy-On-Demand Protocol (How):** Never bulk-copy all seeds into a task directory. Start exclusively with `task.md`. Sequentially instantiate subsequent artifacts (`research.md` $\rightarrow$ `decision.md` $\rightarrow$ `plan.md` $\rightarrow$ `state.md` $\rightarrow$ `review.md` $\rightarrow$ `handoff.md`) only as the task progresses into each phase. In PAIR mode, the agent pauses for human gate approvals; in DELEGATED / Fast-Track mode, the agent auto-instantiates subsequent seeds, auto-certifies gates, and executes continuously without interruptions.
 </architectural_rationale>
 
 ---
@@ -23,7 +23,7 @@
 
   <seed type="task" path="task-template.md.seed">
     Master contract and state record for the full RIPER-5 lifecycle.
-    Contains: Task Control metadata, Spec, Context Boundaries, Verification Strategy,
+    Contains: Task Control metadata (working_mode PAIR/DELEGATED), Spec, Context Boundaries, Verification Strategy,
     Decisions, RIPER-5 Execution Plan (with Gates G0–G3), and Guardrails.
   </seed>
 
@@ -38,7 +38,7 @@
   <seed type="decision" path="decision-template.md.seed">
     Innovate phase artifact. Gate 1 memory.
     Contains: Decision question, Options (A/B/C) with trade-off matrix,
-    Recommendation, Engineer Decision (must be filled by human), Constraints Created,
+    Recommendation, Implementation Decision (filled by engineer in PAIR, or auto-selected in DELEGATED), Constraints Created,
     Evidence Still Required, and Gate 1 checklist.
   </seed>
 
@@ -121,9 +121,13 @@ handoff.md      ← Final projection (short)
 | Execute | source code + tests + `state.md` |
 | Review | `review.md` |
 | Handoff | `handoff.md` |
-| Gate 1 | `decision.md` approval |
-| Gate 2 | `plan.md` approval |
-| Gate 3 | `review.md` approval |
+| Pause / Park | `paused.md` (clean freeze & resumption checklist) |
+| Cancellation | `cancelled.md` (knowledge preservation upon abort) |
+| Benchmark / Metrics | `results.tsv` (performance and eval metrics) |
+| Program | `program.md` (multi-phase umbrella epic blueprint) |
+| Gate 1 | `decision.md` approval (Engineer or Auto-DELEGATED) |
+| Gate 2 | `plan.md` approval (Engineer or Auto-DELEGATED) |
+| Gate 3 | `review.md` approval (Engineer or Auto-DELEGATED) |
 
 ---
 

package/templates/en/process/_seeds/decision-template.md.seed

@@ -2,7 +2,7 @@
 
 <technical_decision task_id="[TASK-ID]" dec_id="[DEC-ID]" version="1.0" framework="RIPER-5">
 
-<!-- READ-ONLY PHASE. Present options. Do not decide without engineer approval. -->
+<!-- READ-ONLY PHASE. Present options. PAIR: engineer decides. DELEGATED: agent auto-selects optimal recommendation and advances. -->
 <decision_status>
   <phase>INNOVATE</phase>
   <mode>READ-ONLY</mode>
@@ -98,9 +98,11 @@
 
 ---
 
-## 6. Engineer Decision
+## 6. Implementation Decision
 
-<!-- This section must be completed by the engineer before Gate 1 passes -->
+<!-- PAIR mode: Completed by engineer before Gate 1 passes.
+     DELEGATED / Fast-Track mode: Agent automatically populates Recommendation
+     into <selected_option>, documents rationale, signs Gate 1 with [AUTO: DELEGATED], and proceeds. -->
 <engineer_decision>
   <selected_option></selected_option>
   <rationale></rationale>
@@ -133,10 +135,9 @@
 
 <gate id="G1">
   - [ ] All options presented and trade-offs analyzed.
-  - [ ] Engineer reviewed options.
-  - [ ] Engineer selected option recorded above.
+  - [ ] Selected option recorded above.
   - [ ] No blocking business / schema / security decision remains open.
-  <approved_by></approved_by>
+  <approved_by></approved_by>  <!-- Engineer name (PAIR) or [AUTO: DELEGATED] (DELEGATED/Fast-Track) -->
   <approved_date></approved_date>
 </gate>
 

package/templates/en/process/_seeds/plan-template.md.seed

@@ -127,8 +127,8 @@
   - [ ] Stop conditions defined per slice.
   - [ ] Rollback point defined per slice.
   - [ ] Allowed commands listed.
-  - [ ] Engineer approved.
-  <approved_by></approved_by>
+  - [ ] Plan approved.
+  <approved_by></approved_by>  <!-- Engineer name (PAIR) or [AUTO: DELEGATED] (DELEGATED/Fast-Track) -->
   <approved_date></approved_date>
 </gate>
 

package/templates/en/process/_seeds/review-template.md.seed

@@ -142,7 +142,7 @@
   - [ ] Residual risk explicitly accepted.
   - [ ] Review decision: PASS.
   - [ ] Ready for handoff.
-  <approved_by></approved_by>
+  <approved_by></approved_by>  <!-- Engineer name (PAIR) or [AUTO: DELEGATED] (DELEGATED/Fast-Track) -->
   <approved_date></approved_date>
 </gate>
 

package/templates/en/process/_seeds/task-template.md.seed

@@ -6,10 +6,12 @@
      SECTION 0 — TASK CONTROL (master state record)
      ════════════════════════════════════════════ -->
 <task_control>
-  <status>BACKLOG</status>  <!-- BACKLOG | ACTIVE | BLOCKED | REVIEW | COMPLETED -->
+  <status>BACKLOG</status>  <!-- BACKLOG | ACTIVE | BLOCKED | PAUSED | REVIEW | COMPLETED | CANCELLED -->
   <spec_level>S1</spec_level>  <!-- S0=sketch | S1=defined | S2=verified | S3=locked -->
+  <priority>P2</priority>  <!-- P0=urgent | P1=high | P2=normal | P3=low -->
   <risk>MEDIUM</risk>  <!-- LOW | MEDIUM | HIGH -->
-  <working_mode>PAIR</working_mode>  <!-- MANUAL | PAIR | DELEGATED | DIAGNOSE-ONLY -->
+  <estimated_story_points>2</estimated_story_points>  <!-- 1 SP ≈ 2-4 focused engineering hours -->
+  <working_mode>PAIR</working_mode>  <!-- PAIR (default: halts at each gate for review) | DELEGATED (fast-track: continuous autonomous run) | MANUAL | DIAGNOSE-ONLY -->
   <current_phase>RESEARCH</current_phase>  <!-- RESEARCH | INNOVATE | PLAN | EXECUTE | REVIEW -->
   <owner>@engineer</owner>
   <decision_owner>@engineer</decision_owner>
@@ -63,6 +65,11 @@
     - `tests/[path/to/test_file]` — [Target assertions and test fixtures]
   </target_files>
 
+  <context_groups>
+    <!-- References to registered context groups from process/context/all-context.md -->
+    - [planning | tests | protocols | domain-specific-group]
+  </context_groups>
+
   <source_of_truth>
     <requirement>[Link to requirement or spec]</requirement>
     <architecture>[Link to architecture doc or ADR]</architecture>
@@ -117,18 +124,19 @@
       - [ ] Current behavior understood and documented.
       - [ ] Execution flow traced.
       - [ ] No unresolved research blocker.
+      <!-- In DELEGATED / Fast-Track mode: Agent auto-certifies and advances to Innovate immediately -->
     </gate>
   </phase>
 
   <phase name="Innovate" order="2">
-    <!-- READ-ONLY. Present 2-3 viable options. Do not decide unilaterally. -->
+    <!-- READ-ONLY. Present 2-3 viable options. PAIR: engineer selects. DELEGATED: agent auto-selects optimal option. -->
     - [ ] Generate 2–3 alternative approaches with trade-off matrix.
     - [ ] Produce `decision.md` artifact.
     <gate id="G1" label="Gate 1 — Decision Approved">
-      - [ ] Options reviewed by engineer.
+      - [ ] Options reviewed and trade-offs analyzed.
       - [ ] Selected option recorded in `decision.md`.
       - [ ] No blocking business/schema/security decision remains open.
-      <approved_by></approved_by>
+      <approved_by></approved_by>  <!-- Engineer name (PAIR) or [AUTO: DELEGATED] (DELEGATED/Fast-Track) -->
       <approved_date></approved_date>
     </gate>
   </phase>
@@ -142,8 +150,8 @@
       - [ ] Allowed/forbidden file scope is defined.
       - [ ] Rollback point defined per slice.
       - [ ] Stop conditions defined.
-      - [ ] Engineer approved.
-      <approved_by></approved_by>
+      - [ ] Plan approved.
+      <approved_by></approved_by>  <!-- Engineer name (PAIR) or [AUTO: DELEGATED] (DELEGATED/Fast-Track) -->
       <approved_date></approved_date>
     </gate>
   </phase>
@@ -154,6 +162,7 @@
     - [ ] Run verifier after each slice.
     - [ ] Inspect diff after each slice.
     - [ ] Update `state.md` after each slice.
+    - [ ] (Optional) Record quantitative metric deltas in `results.tsv` if tracking benchmarks/performance.
   </phase>
 
   <phase name="Review" order="5">
@@ -163,9 +172,9 @@
     <gate id="G3" label="Gate 3 — Review Passed">
       - [ ] All AC verified with evidence.
       - [ ] Residual risk accepted.
-      - [ ] Review decision: PASS | CHANGES_REQUIRED | BLOCKED.
+      - [ ] Review decision: PASS.
       - [ ] Ready for handoff.
-      <approved_by></approved_by>
+      <approved_by></approved_by>  <!-- Engineer name (PAIR) or [AUTO: DELEGATED] (DELEGATED/Fast-Track) -->
       <approved_date></approved_date>
     </gate>
   </phase>

package/templates/en/process/context/all-context.md

@@ -26,7 +26,7 @@
   <group id="seeds">
     <title>Seeds & Archetypes Guide</title>
     <path>[`../_seeds/_GUIDE.md`](../_seeds/_GUIDE.md)</path>
-    <scope>Scaffolding seeds for tasks, research, decisions, plans, state, reviews, handoffs, and programs. Full artifact chain and instantiation commands.</scope>
+    <scope>Scaffolding seeds for tasks, research, decisions, plans, state, reviews, handoffs, pause, cancellation, benchmarks, and programs. Full artifact chain and instantiation commands.</scope>
   </group>
 
   <group id="tests">

package/templates/en/process/context/planning/all-planning.md

@@ -1,31 +1,87 @@
 # Planning Standards & Calibration Context
 
-<planning_context version="1.0">
+<planning_context version="2.0">
 
 <overview>
-  Guidelines for story point estimation, vertical slicing, and capacity calibration.
+  Engineering guidelines for story point estimation, vertical slicing, rollback checkpointing, and capacity calibration.
 </overview>
 
-## 1. Quality Standards
+---
+
+## 1. Quality Standards (INVEST Criteria)
+
 <quality_standards>
   <standard name="INVEST">
-    - **Independent:** Deliverable without cross-story blockers.
-    - **Negotiable:** Implementation details are flexibly refined.
-    - **Valuable:** Delivers measurable user or business value.
-    - **Estimable:** Scoped clearly to estimate effort.
-    - **Small:** Fits within 1-3 engineering days.
-    - **Testable:** Concrete Acceptance Criteria defined.
-  </standard>
-
-  <standard name="Vertical Slicing">
-    Avoid horizontal silos. Deliver end-to-end vertical slices across API, logic, and persistence.
+    - **Independent:** Deliverable and verifiable without cross-story circular blockers.
+    - **Negotiable:** Implementation details and technical approaches can be flexibly evaluated during the Innovate phase.
+    - **Valuable:** Delivers demonstrable capability or verifiable engineering progress.
+    - **Estimable:** Scoped tightly enough that effort and file touchpoints can be calculated.
+    - **Small:** Sized to fit comfortably within 1–3 engineering days (or 1–3 Story Points).
+    - **Testable:** Accompanied by unambiguous, verifiable Acceptance Criteria (`- [ ]`).
   </standard>
 </quality_standards>
 
-## 2. Capacity Calibration
+---
+
+## 2. Vertical Slicing Discipline
+
+<vertical_slicing>
+  ### Vertical Slice vs Horizontal Silo
+  - **The Vertical Slice Rule:** Each slice must cut through all required technical layers (e.g. Domain Model $\rightarrow$ Service/Logic $\rightarrow$ Port Adapter/API $\rightarrow$ Automated Test) to deliver one complete, verifiable increment of behavior.
+  - **The Horizontal Silo Anti-Pattern:** Never slice horizontally (e.g., Slice 1: create all database tables, Slice 2: create all services, Slice 3: create all endpoints). Horizontal layers cannot be verified independently, leaving the system in a broken or unprovable state until the final slice.
+
+  ### Vertical Slicing Workflow
+  ```text
+  [Slice 1: Thin End-to-End Core] ──► Domain Entity + In-Memory Repo + Minimal Endpoint + Passing Test
+  [Slice 2: Persistence & Boundary] ──► Real Database Adapter + Migration + Integration Verifier
+  [Slice 3: Edge Cases & Validation] ──► Strict Input Sanitization + Error Handlers + Boundary Tests
+  ```
+</vertical_slicing>
+
+---
+
+## 3. Slice Sizing, Atomic Verifiability & Checkpoints
+
+<slice_sizing>
+  <rule id="slice_cardinality">
+    A standard task should contain between **2 and 5 vertical slices**. If a task requires more than 5 slices, it has high blast radius and must be decomposed into a Multi-Phase Program (`program-template.md.seed`).
+  </rule>
+
+  <rule id="autonomous_verifier">
+    Every single slice in `plan.md` must have an explicit, executable `<verifier>` command (e.g. `npm run test -- test/path/test.js` or `pytest tests/unit/test_slice.py`). A slice is NOT complete until its verifier exits with code 0.
+  </rule>
+
+  <rule id="atomic_checkpoint">
+    Commit or checkpoint each slice atomically upon verification passing. This guarantees clean git history and enables instant rollback if subsequent slices encounter irrecoverable regressions.
+  </rule>
+</slice_sizing>
+
+---
+
+## 4. Rollback Strategy per Slice
+
+<rollback_strategy>
+  Every slice in `plan.md` MUST specify a concrete `<rollback_point>`:
+  - **Git Checkpoint:** `git checkout -- <allowed_files>` or revert commit hash.
+  - **Stash Checkpoint:** `git stash pop` or dedicated WIP branch.
+  - **Data/Schema Rollback:** Down-migration script or drop ephemeral test container.
+
+  If the agent exhausts its 3-attempt retry budget on a slice, it must execute the rollback point before halting and escalating to human guidance.
+</rollback_strategy>
+
+---
+
+## 5. Story Point Capacity Calibration
+
 <capacity_calibration>
-  <unit>1 Story Point ≈ 2-4 focused engineering hours</unit>
-  <max_task_size>3-5 Story Points (larger tasks must be decomposed)</max_task_size>
+  <unit>1 Story Point (SP) ≈ 2–4 focused engineering hours</unit>
+  <scale>
+    - **1 SP:** Straightforward change with clear scope (1-3 files touched, 1-2 slices).
+    - **2 SP:** Standard feature or refactoring (3-5 files touched, 2-3 slices).
+    - **3 SP:** Moderately complex task touching domain contracts and persistence (4-7 files, 3-4 slices).
+    - **5 SP:** Maximum size for a single active task in `process/features/active/`.
+    - **> 5 SP:** Oversized. Must be split into multiple standalone tasks or organized under `program-template.md.seed`.
+  </scale>
 </capacity_calibration>
 
 </planning_context>