opencode-orchestrator 1.0.34 → 1.0.36

package/README.md

@@ -23,50 +23,17 @@ In an OpenCode environment:
 
 ## Overview
 
-OpenCode Orchestrator is a framework designed to manage complex software engineering tasks through parallel multi-agent execution. It extends the capabilities of standard AI agents by introducing a hierarchical delegation model and a multi-stage verification pipeline.
-
-### Key Characteristics
-*   **Parallel Execution**: Manages multiple concurrent agent sessions to accelerate development.
-*   **Hierarchical Delegation**: Supports recursive task decomposition (Fractal Spawning) for complex requirements.
-*   **Integrated Verification**: Employs a multi-stage pipeline (Unit & Integration) to ensure code quality.
-*   **Persistent State**: Uses structured logging to maintain state and recover from interruptions.
-
----
-
-## Core Philosophy: Adaptive Engineering
-
-The orchestrator follows an **Explore → Learn → Adapt → Act** cycle to ensure agents remain grounded in the project's specific context.
-
-*   **Explore**: Systematically discover the project structure, tech stack, and documentation.
-*   **Learn**: Extract existing patterns, naming conventions, and architectural decisions from the codebase.
-*   **Adapt**: Adjust implementation strategies and verification parameters based on learned context.
-*   **Act**: Execute tasks through parallel delegation with evidence-based verification.
-
----
-
-## 🏗️ Architecture
-
-The system is built upon two core mechanisms that handle scaling and quality control.
-
-### HPFA (Hyper-Parallel Fractal Architecture)
-HPFA is a parallel execution model that enables task decomposition at scale.
-*   **Managed Concurrency**: Orchestrates up to 50 parallel agent sessions simultaneously.
-*   **Recursive Scaling**: Allows "Worker" agents to spawn sub-workers for modular tasks, ensuring deep architectural coverage.
-
-### MSVP (Multi-Stage Verification Pipeline)
-MSVP is a structured verification process that decouples implementation from quality assurance.
-*   **Stage 1 (Unit Verification)**: Reviewers validate individual module changes and run local tests immediately after implementation.
-*   **Stage 2 (Integration Review)**: A master reviewer verifies cross-module consistency and system integrity after all individual units are completed.
+OpenCode Orchestrator manages complex software tasks through **parallel multi-agent execution**. Commander orchestrates Workers and Reviewers to implement and verify code concurrently.
 
 ---
 
-## 📊 Workflow Diagram
+## 📊 Workflow
 
 ```text
               [User Task Input]
                      │
            ┌─────────▼─────────┐
-           │     COMMANDER     │  (Orchestration & Discovery)
+           │     COMMANDER     │  (Orchestration)
            └─────────┬─────────┘
                      │
            ┌─────────▼─────────┐
@@ -78,7 +45,7 @@ MSVP is a structured verification process that decouples implementation from qua
     └──────┬─────────┬─────────┬──────┘
            │         │         │
     ┌──────▼──┐ ┌────▼───┐ ┌───▼────┐
-    │ WORKER A│ │ WORKER B│ │ WORKER C│  (Implementation)
+    │ WORKER  │ │ WORKER │ │ WORKER │
     └──────┬──┘ └────┬───┘ └────┬────┘
            │         │          │
     ╔══════▼═════════▼══════════▼══════╗
@@ -86,44 +53,39 @@ MSVP is a structured verification process that decouples implementation from qua
     ╚══════╤═════════╤══════════╤══════╝
            │         │          │
     ┌──────▼──┐ ┌────▼───┐ ┌────▼────┐
-    │REVIEWER │ │REVIEWER │ │REVIEWER │  (Unit Verification)
+    │REVIEWER │ │REVIEWER │ │REVIEWER │
     └──────┬──┘ └────┬───┘ └────┬────┘
            │         │          │
-    ═══════▼═════════▼══════════▼══════
-    │          SYNC BARRIER           │
-    ════════════════╤══════════════════
-                    │
-          ┌─────────▼─────────┐
-          │  MASTER REVIEWER  │  (E2E Verification)
-          └─────────┬─────────┘
-                    │
-          ┌─────────▼─────────┐
-          │  Mission Sealed?  │
-          └─────────┬─────────┘
-               No ↙   ↘ Yes
-            [Loop]   [Complete]
+          ═▼═════════▼══════════▼═
+          │     SYNC BARRIER     │
+          ═══════════╤═══════════
+                     │
+           ┌─────────▼─────────┐
+           │  MASTER REVIEWER  │  (E2E Verification)
+           └─────────┬─────────┘
+                     │
+                [MISSION SEALED]
 ```
 
 ---
 
-## 🚀 System Roles
+## 🚀 Agents
 
-| Role | Responsibility | Key Actions |
-|:-----|:---------------|:------------|
-| **Commander** | Mission Control | Task assignment, conflict resolution, global synchronization |
-| **Planner** | Architecture | Environment analysis, dependency mapping, TODO generation |
-| **Worker** | Implementation | Code writing, file modification, unit test creation |
-| **Reviewer** | Quality Control| Static analysis, build verification, integration testing |
+| Agent | Role |
+|:------|:-----|
+| **Commander** | Orchestrates all agents, manages task flow |
+| **Planner** | Creates TODO.md with task breakdown |
+| **Worker** | Implements features, writes tests |
+| **Reviewer** | Validates code, runs verification |
 
 ---
 
-## 🛠️ Features
+## ✨ Key Features
 
-*   **Concurrent Task Management**: Efficiently handles up to 50 background agent sessions.
-*   **Automated Context Synthesis**: Parallel scouters gather environment intelligence (Structure, Stack, Docs) instantly.
-*   **Synchronized Verification**: Ensures all distributed tasks pass Stage 1 review before final integration.
-*   **Fault Tolerance**: Automatically resumes progress from checkpoints in case of tool or session failure.
-*   **Context Optimization**: Monitors context window limits and performs automated compaction for long sessions.
+- **Parallel Execution**: Up to 50 concurrent agent sessions
+- **Two-Stage Verification**: Unit review → Master review → Seal
+- **Fault Tolerance**: Auto-recovery from failures
+- **Context Optimization**: Manages token limits automatically
 
 ---
 

package/dist/agents/prompts/reviewer/forbidden.d.ts

@@ -2,5 +2,6 @@
  * Reviewer Forbidden Actions
  *
  * Verification integrity - never approve without evidence.
+ * Reviewer is a TERMINAL node and cannot spawn other agents.
  */
 export declare const REVIEWER_FORBIDDEN: string;

package/dist/agents/prompts/worker/forbidden.d.ts

@@ -2,5 +2,6 @@
  * Worker Forbidden Actions
  *
  * Universal anti-patterns - adapt to project-specific conventions.
+ * Worker is a TERMINAL node and cannot spawn other agents.
  */
 export declare const WORKER_FORBIDDEN: string;

package/dist/index.js

@@ -990,6 +990,30 @@ var SCOUT_INTEGRATION = {
   PROMPT: "Read all modified files and their cross-module dependencies in parallel. Map the actual integration surface and identify potential consistency issues."
 };
 
+// src/shared/prompt/constants/architecture.ts
+var AGENT_ROLES = {
+  /** Only Commander can spawn other agents */
+  SPAWNER: AGENT_NAMES.COMMANDER,
+  /** Terminal agents cannot spawn any sub-agents */
+  TERMINAL_AGENTS: [AGENT_NAMES.PLANNER, AGENT_NAMES.WORKER, AGENT_NAMES.REVIEWER]
+};
+var TERMINAL_NODE = {
+  /** Label for terminal node concept */
+  LABEL: "TERMINAL node",
+  /** Reason why terminal nodes can't spawn */
+  REASON: "prevents infinite recursion",
+  /** What to do instead of spawning */
+  ALTERNATIVE: `Report back to ${AGENT_NAMES.COMMANDER} with specific blockers`
+};
+var SPAWNING_RULES = {
+  /** Who can spawn */
+  SPAWNER_ONLY: `${AGENT_NAMES.COMMANDER} is the ONLY agent who can spawn other agents`,
+  /** What terminal nodes should do */
+  TERMINAL_BEHAVIOR: `Complete your assigned task directly without delegation`,
+  /** Error message for blocked spawn attempts */
+  BLOCKED_MESSAGE: `You are a ${TERMINAL_NODE.LABEL} - ${TERMINAL_NODE.REASON}`
+};
+
 // src/core/agents/consts/task-status.const.ts
 var TASK_STATUS = {
   PENDING: STATUS_LABEL.PENDING,
@@ -14298,8 +14322,8 @@ var PLANNER_FORBIDDEN = `${PROMPT_TAGS.FORBIDDEN_ACTIONS.open}
 ## \u26D4 NEVER Spawn or Delegate (CRITICAL)
 - NEVER use \`${TOOL_NAMES.DELEGATE_TASK}\` to spawn agents
 - NEVER use \`${TOOL_NAMES.CALL_AGENT}\` to create sessions
-- You are a TERMINAL node - create plans, don't execute them
-- ${AGENT_NAMES.COMMANDER} is the ONLY agent who can spawn Workers/Reviewers
+- You are a ${TERMINAL_NODE.LABEL} - create plans, don't execute them
+- ${SPAWNING_RULES.SPAWNER_ONLY}
 
 ## Never Implement
 - NEVER write actual code \u2192 Only plan and research
@@ -14608,9 +14632,9 @@ var WORKER_FORBIDDEN = `${PROMPT_TAGS.FORBIDDEN_ACTIONS.open}
 ## \u26D4 NEVER Spawn or Delegate (CRITICAL)
 - NEVER use \`${TOOL_NAMES.DELEGATE_TASK}\` to spawn sub-workers
 - NEVER use \`${TOOL_NAMES.CALL_AGENT}\` to create additional sessions
-- You are a TERMINAL node - complete your assigned file directly
-- If task is too complex, REPORT BACK to ${AGENT_NAMES.COMMANDER} with specific blockers
-- Violating this rule causes infinite recursion and system failure
+- You are a ${TERMINAL_NODE.LABEL} - ${SPAWNING_RULES.TERMINAL_BEHAVIOR}
+- If task is too complex, ${TERMINAL_NODE.ALTERNATIVE}
+- Violating this rule ${TERMINAL_NODE.REASON}
 
 ## Never Assume
 - NEVER guess API syntax \u2192 CHECK ${PATHS.DOCS}/ or research first
@@ -15036,9 +15060,9 @@ var REVIEWER_FORBIDDEN = `${PROMPT_TAGS.FORBIDDEN_ACTIONS.open}
 ## \u26D4 NEVER Spawn or Delegate (CRITICAL)
 - NEVER use \`${TOOL_NAMES.DELEGATE_TASK}\` to spawn additional reviewers
 - NEVER use \`${TOOL_NAMES.CALL_AGENT}\` to create sub-sessions
-- You are a TERMINAL node - verify your assigned file directly
-- If verification scope is too large, REPORT BACK to ${AGENT_NAMES.COMMANDER}
-- Violating this rule causes infinite recursion and blocks Master Review
+- You are a ${TERMINAL_NODE.LABEL} - verify your assigned file directly
+- If verification scope is too large, ${TERMINAL_NODE.ALTERNATIVE}
+- Violating this rule ${TERMINAL_NODE.REASON}
 
 ## Never Approve Without Verification
 - NEVER approve without actually running the project's test command

package/dist/shared/prompt/constants/architecture.d.ts

@@ -0,0 +1,64 @@
+/**
+ * Architecture Rules - Core system behavior constants
+ *
+ * These define the immutable rules about agent hierarchy and spawning.
+ * Used by prompts to ensure consistent messaging across all agents.
+ */
+/**
+ * Agent hierarchy roles
+ */
+export declare const AGENT_ROLES: {
+    /** Only Commander can spawn other agents */
+    readonly SPAWNER: "Commander";
+    /** Terminal agents cannot spawn any sub-agents */
+    readonly TERMINAL_AGENTS: readonly ["Planner", "Worker", "Reviewer"];
+};
+/**
+ * Terminal node messaging - used in forbidden actions
+ */
+export declare const TERMINAL_NODE: {
+    /** Label for terminal node concept */
+    readonly LABEL: "TERMINAL node";
+    /** Reason why terminal nodes can't spawn */
+    readonly REASON: "prevents infinite recursion";
+    /** What to do instead of spawning */
+    readonly ALTERNATIVE: "Report back to Commander with specific blockers";
+};
+/**
+ * Spawning restriction messages
+ */
+export declare const SPAWNING_RULES: {
+    /** Who can spawn */
+    readonly SPAWNER_ONLY: "Commander is the ONLY agent who can spawn other agents";
+    /** What terminal nodes should do */
+    readonly TERMINAL_BEHAVIOR: "Complete your assigned task directly without delegation";
+    /** Error message for blocked spawn attempts */
+    readonly BLOCKED_MESSAGE: "You are a TERMINAL node - prevents infinite recursion";
+};
+/**
+ * Core architecture rules as a single exportable constant
+ */
+export declare const ARCHITECTURE_RULES: {
+    readonly AGENT_ROLES: {
+        /** Only Commander can spawn other agents */
+        readonly SPAWNER: "Commander";
+        /** Terminal agents cannot spawn any sub-agents */
+        readonly TERMINAL_AGENTS: readonly ["Planner", "Worker", "Reviewer"];
+    };
+    readonly TERMINAL_NODE: {
+        /** Label for terminal node concept */
+        readonly LABEL: "TERMINAL node";
+        /** Reason why terminal nodes can't spawn */
+        readonly REASON: "prevents infinite recursion";
+        /** What to do instead of spawning */
+        readonly ALTERNATIVE: "Report back to Commander with specific blockers";
+    };
+    readonly SPAWNING_RULES: {
+        /** Who can spawn */
+        readonly SPAWNER_ONLY: "Commander is the ONLY agent who can spawn other agents";
+        /** What terminal nodes should do */
+        readonly TERMINAL_BEHAVIOR: "Complete your assigned task directly without delegation";
+        /** Error message for blocked spawn attempts */
+        readonly BLOCKED_MESSAGE: "You are a TERMINAL node - prevents infinite recursion";
+    };
+};

package/dist/shared/prompt/constants/index.d.ts

@@ -6,3 +6,4 @@ export { WORK_STATUS } from "./status.js";
 export { PHILOSOPHY_PHASES, PHILOSOPHY_TAGLINE, PHILOSOPHY_QUOTE, PHILOSOPHY_LEARN_PRINCIPLE, EXECUTION_CYCLE, EXECUTION_CYCLE_STEPS, } from "./philosophy.js";
 export { PHASE_0_DIRECT_DISCOVERY, PHASE_1_THINK_ANALYSIS, PHASE_5_MSVP, HPFA_RULES } from "./mandates.js";
 export { SCOUT_LABEL, SCOUT_STRUCTURE, SCOUT_STACK, SCOUT_DOCS, SCOUT_INFRA, SCOUT_INTEGRATION } from "./scouts.js";
+export { AGENT_ROLES, TERMINAL_NODE, SPAWNING_RULES, ARCHITECTURE_RULES, } from "./architecture.js";

package/package.json

@@ -2,7 +2,7 @@
   "name": "opencode-orchestrator",
   "displayName": "OpenCode Orchestrator",
   "description": "Distributed Cognitive Architecture for OpenCode. Turns simple prompts into specialized multi-agent workflows (Planner, Coder, Reviewer).",
-  "version": "1.0.34",
+  "version": "1.0.36",
   "author": "agnusdei1207",
   "license": "MIT",
   "repository": {