opencode-multiagent 0.3.0-next.1 → 0.5.0

package/docs/usage-guide.tr.md

@@ -33,42 +33,43 @@ OpenCode paketi yüklediğinde plugin şu adımları uygular:
 3. efektif runtime flag ve agent ayarlarını oluşturur
 4. paketlenmiş agent ve command markdown dosyalarını aktif config'e derler
 5. host config içinde yoksa plugin'e ait MCP varsayılanlarını ekler
-6. mümkünse `lead` ajanını varsayılan top-level ajan olarak tercih eder
-7. telemetry, file lock, supervision, quality reminder ve ortak task board hook'larını açar
+6. mümkünse `planner` ajanını varsayılan top-level ajan olarak tercih eder
+7. telemetry, supervision, quality reminder ve ortak task board hook'larını açar
 
 ## 3. Normal Kullanım Akışı
 
-Normal kullanımda worker ajanları elle seçmeniz gerekmez. Sistem, top-level etkileşimin `lead`
-üzerinden gitmesi için tasarlanmıştır.
+Normal kullanımda uzman ajanları elle seçmeniz gerekmez. Sistem, çoğu görev için top-level
+etkileşimin `planner` üzerinden gitmesi için tasarlanmıştır. Yönün belirsiz olduğu keşif
+çalışmalarında önce `brainstormer` ile başlayabilirsiniz.
 
-Tipik akış:
+Tipik akışlar:
 
 ```text
-kullanici -> lead -> critic/planner/executor -> uzman alt ajanlar
+kullanici -> planner -> executor -> uzman alt ajanlar
+kullanici -> brainstormer -> (brief ile sonuçlanır) -> planner -> executor -> ...
 ```
 
 ### Temel routing rolleri
 
-| Agent      | Rol                                                                        |
-| ---------- | -------------------------------------------------------------------------- |
-| `lead`     | Talebin tüm yaşam döngüsünü yönetir, route seçer ve sonuçları birleştirir. |
-| `critic`   | İş başlamadan önce rota veya repo durumunu sorgular.                       |
-| `planner`  | Büyük veya riskli işlerde kalıcı bir yürütme planı üretir.                 |
-| `executor` | Planı bounded worker görevlerine bölerek uygular ve sonuçları doğrular.    |
+| Agent          | Rol                                                                                    |
+| -------------- | -------------------------------------------------------------------------------------- |
+| `brainstormer` | Planlama öncesi kullanıcıyla fikirleri keşfeder, planner'a hazır brief üretir.        |
+| `planner`      | Talebi triage eder, route seçer, büyük işler için plan üretir, sonuçları inceler.     |
+| `executor`     | Planı bounded uzman görevlerine bölerek uygular ve sonuçları doğrular.                |
 
-## 4. Triage Seviyeleri
+## 4. Triage Seviyeleri (Tier Modeli)
 
-`lead`, katmanlı bir routing modeli kullanır.
+`planner`, katmanlı bir routing modeli kullanır.
 
-| Seviye   | Anlamı                                                         | Tipik route                             |
-| -------- | -------------------------------------------------------------- | --------------------------------------- |
-| `Tier 0` | Küçük, açık, düşük riskli işler                                | `lead -> executor -> quick/worker`      |
-| `Tier 1` | Kısa bir challenge turundan fayda gören bounded işler          | `lead -> critic -> executor`            |
-| `Tier 2` | Cross-cutting, riskli, belirsiz veya migration ağırlıklı işler | `lead -> critic -> planner -> executor` |
-| `Tier 3` | İnceleme ve repo-memory işleri                                 | `lead -> critic`                        |
+| Tier     | Açıklama                                       | Rota                                          |
+| -------- | ---------------------------------------------- | --------------------------------------------- |
+| `Tier 0` | Küçük, açık işler                              | `planner -> executor -> coder`                |
+| `Tier 1` | Bounded işler                                  | `planner -> executor -> coder/ui-coder/sec-coder` |
+| `Tier 2` | Karmaşık çok modüllü işler                     | `planner -> auditor -> executor -> coders`    |
+| `Tier 3` | İnceleme                                       | `planner` (inspeksiyon modu)                  |
 
 Kullanıcı açısından pratik kural basittir: güçlü bir sebep yoksa görevi doğal şekilde tarif edin ve
-`lead` ajanının route seçmesine izin verin.
+`planner` ajanının route seçmesine izin verin.
 
 ## 5. Ortak Task Board
 
@@ -78,7 +79,7 @@ Plugin üç ortak task aracı sunar:
 - `task_update`
 - `task_list`
 
-Bu araçlar ajan sistemi tarafından dahili olarak kullanılır; ancak log ve `.magent` çıktılarının
+Bu araçlar ajan sistemi tarafından dahili olarak kullanılır; ancak log ve `.opencode` çıktılarının
 nasıl oluştuğunu anlamak için bilinmeleri faydalıdır.
 
 ### Task yaşam döngüsü
@@ -96,7 +97,7 @@ Olası task durumları:
 
 Bir proje kökü varsa task board şu dosyaya yazılır:
 
-- `.magent/tasks/taskboard.json`
+- `.opencode/tasks/taskboard.json`
 
 Bu sayede çok adımlı yürütmeler oturum sonrasında da incelenebilir.
 
@@ -130,11 +131,11 @@ kullanın.
 ```json
 {
   "agentSettings": {
-    "lead": {
-      "steps": 650
+    "planner": {
+      "steps": 300
     },
     "reviewer": {
-      "model": "openai/gpt-5.4"
+      "model": "anthropic/claude-opus-4-6"
     }
   }
 }
@@ -149,8 +150,8 @@ Belirli bir OpenCode alanını açıkça override etmek istediğinizde `opencode
   "$schema": "https://opencode.ai/config.json",
   "plugin": ["opencode-multiagent"],
   "agent": {
-    "lead": {
-      "steps": 650
+    "planner": {
+      "steps": 300
     }
   }
 }
@@ -198,13 +199,13 @@ Bunu açıkça etkinleştirmek için `opencode.json` kullanın:
 
 Sık görülen dosya ve dizinler:
 
-| Yol                                                   | Amaç                                                      |
-| ----------------------------------------------------- | --------------------------------------------------------- |
-| `~/.config/opencode/logs/opencode-multiagent.jsonl`   | Observation ve telemetry logu                             |
-| `~/.config/opencode/plugins/opencode-multiagent.json` | Kullanıcı runtime override dosyası                        |
-| `.magent/tasks/taskboard.json`                        | Ortak task board verisi                                   |
-| `.magent/plans/`                                      | Planner ve scribe tarafından yazılan kalıcı planlar       |
-| `.magent/exec/`                                       | Büyük yürütmeler sırasında tutulan execution artifaktları |
+| Yol                                                   | Amaç                                                           |
+| ----------------------------------------------------- | -------------------------------------------------------------- |
+| `~/.config/opencode/logs/opencode-multiagent.jsonl`   | Observation ve telemetry logu                                  |
+| `~/.config/opencode/plugins/opencode-multiagent.json` | Kullanıcı runtime override dosyası                             |
+| `.opencode/tasks/taskboard.json`                      | Ortak task board verisi                                        |
+| `.opencode/plans/`                                    | `planner` ve `docmaster` tarafından yazılan kalıcı planlar     |
+| `.opencode/exec/`                                     | Büyük yürütmeler sırasında tutulan execution artifaktları      |
 
 ## 10. Geliştirme Akışı
 

package/examples/opencode.with-overrides.json

@@ -12,8 +12,8 @@
     }
   },
   "agent": {
-    "lead": {
-      "steps": 650
+    "planner": {
+      "steps": 300
     }
   }
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "opencode-multiagent",
-  "version": "0.3.0-next.1",
+  "version": "0.5.0",
   "type": "module",
   "description": "Multi-agent orchestration plugin for OpenCode",
   "license": "MIT",

package/skills/AGENTS.md

@@ -0,0 +1,51 @@
+# skills/
+
+Bundled skill definitions for the multi-agent system. Each skill is a reusable prompt injection loaded by agents at runtime.
+
+## Directory Convention
+
+Each skill is a subdirectory containing exactly two files:
+
+- `manifest.json` — metadata and routing configuration
+- `SKILL.md` — the skill content injected into agent context
+
+No other files should be added to a skill directory.
+
+## manifest.json Schema
+
+| Field                | Type     | Required | Description                                                          |
+| -------------------- | -------- | -------- | -------------------------------------------------------------------- |
+| `name`               | string   | yes      | Skill identifier, must match the directory name                      |
+| `version`            | string   | yes      | Semver version                                                       |
+| `description`        | string   | yes      | One-line description of what the skill provides                      |
+| `triggers`           | string[] | yes      | Phrases that activate this skill (used for matching)                 |
+| `applicable_agents`  | string[] | yes      | Which agents may use this skill (must match agent permission blocks) |
+| `max_context_tokens` | number   | yes      | Token budget for injection — keep as low as practical                |
+| `entry_file`         | string   | yes      | Always `"SKILL.md"`                                                  |
+
+## Naming Prefixes
+
+| Prefix       | Domain                           | Examples                                     |
+| ------------ | -------------------------------- | -------------------------------------------- |
+| `cek-`       | Context and prompt engineering   | `cek-context-engineering`, `cek-test-prompt` |
+| `sadd-`      | Structured agent debate/decision | `sadd-judge-with-debate`                     |
+| `reflexion-` | Reflection and critique patterns | `reflexion-reflect`, `reflexion-critique`    |
+| (none)       | General-purpose skills           | `executing-plans`, `task-decomposition`      |
+
+## applicable_agents Coupling
+
+The `applicable_agents` array must match the `skill:` permission block in the corresponding agent definition files (`agents/*.md`). If an agent is listed in `applicable_agents` but does not have `skill: <skill-name>: allow` in its permission block, the skill will not be loadable by that agent.
+
+## Skill Resolution Order
+
+At runtime, skills are resolved from multiple sources in this order:
+
+1. `${plugin_root}/skills` — bundled skills (this directory)
+2. `~/.agents/skills` — user global skills
+3. `~/skills` — user local skills
+
+Later sources can override earlier ones by using the same directory name.
+
+## skill_injection Flag
+
+The `skill_injection` flag in plugin settings controls whether skills are dynamically injected into agent prompts. It is currently `false` by default (reserved for future use). Skills are still loadable via the `skill` tool regardless of this flag.

package/skills/advanced-evaluation/manifest.json

@@ -3,7 +3,7 @@
   "version": "1.0.0",
   "description": "Advanced evaluation workflows for comparative and bias-aware judgment tasks",
   "triggers": ["advanced evaluation", "compare outputs", "pairwise", "position bias", "judge"],
-  "applicable_agents": ["librarian"],
+  "applicable_agents": ["scout"],
   "max_context_tokens": 2200,
   "entry_file": "SKILL.md"
 }

package/skills/cek-context-engineering/manifest.json

@@ -9,7 +9,7 @@
     "lost in middle",
     "context design"
   ],
-  "applicable_agents": ["critic"],
+  "applicable_agents": ["planner"],
   "max_context_tokens": 2600,
   "entry_file": "SKILL.md"
 }

package/skills/cek-prompt-engineering/manifest.json

@@ -9,7 +9,7 @@
     "rewrite prompt",
     "instruction design"
   ],
-  "applicable_agents": ["critic"],
+  "applicable_agents": ["planner"],
   "max_context_tokens": 2200,
   "entry_file": "SKILL.md"
 }

package/skills/cek-test-prompt/manifest.json

@@ -9,7 +9,7 @@
     "prompt behavior",
     "red green prompt"
   ],
-  "applicable_agents": ["critic"],
+  "applicable_agents": ["planner"],
   "max_context_tokens": 2400,
   "entry_file": "SKILL.md"
 }

package/skills/cek-thought-based-reasoning/manifest.json

@@ -9,7 +9,7 @@
     "tree of thoughts",
     "reasoning"
   ],
-  "applicable_agents": ["critic"],
+  "applicable_agents": ["planner"],
   "max_context_tokens": 2600,
   "entry_file": "SKILL.md"
 }

package/skills/context-degradation/manifest.json

@@ -9,7 +9,7 @@
     "context confusion",
     "attention pattern"
   ],
-  "applicable_agents": ["strategist"],
+  "applicable_agents": ["planner"],
   "max_context_tokens": 2200,
   "entry_file": "SKILL.md"
 }

package/skills/debate/manifest.json

@@ -3,7 +3,7 @@
   "version": "1.0.0",
   "description": "Structured debate templates for stress-testing positions and solutions",
   "triggers": ["debate", "argue", "stress test", "devils advocate", "counter argument"],
-  "applicable_agents": ["critic", "strategist", "librarian"],
+  "applicable_agents": ["planner", "scout"],
   "max_context_tokens": 2200,
   "entry_file": "SKILL.md"
 }

package/skills/design-first/manifest.json

@@ -3,7 +3,7 @@
   "version": "1.0.0",
   "description": "Design the approach before implementation starts",
   "triggers": ["design", "architecture", "plan", "schema", "api", "interface"],
-  "applicable_agents": ["planner", "strategist", "heavy-worker", "deep-worker", "advisor"],
+  "applicable_agents": ["planner", "sec-coder"],
   "max_context_tokens": 1500,
   "entry_file": "SKILL.md"
 }

package/skills/dispatching-parallel-agents/manifest.json

@@ -9,7 +9,7 @@
     "independent subtasks",
     "parallelize"
   ],
-  "applicable_agents": ["critic", "strategist"],
+  "applicable_agents": ["planner"],
   "max_context_tokens": 1800,
   "entry_file": "SKILL.md"
 }

package/skills/drift-analysis/manifest.json

@@ -3,7 +3,7 @@
   "version": "1.0.0",
   "description": "Analyze plan drift between documented intent and repository reality",
   "triggers": ["drift", "reality check", "plan drift", "implementation gap", "roadmap alignment"],
-  "applicable_agents": ["planner", "auditor", "strategist"],
+  "applicable_agents": ["planner", "auditor"],
   "max_context_tokens": 2000,
   "entry_file": "SKILL.md"
 }

package/skills/evaluation/manifest.json

@@ -3,7 +3,7 @@
   "version": "1.0.0",
   "description": "Lightweight evaluation framework for judgments and comparisons",
   "triggers": ["evaluate", "assess", "judge", "compare", "rate"],
-  "applicable_agents": ["reviewer", "librarian"],
+  "applicable_agents": ["reviewer", "scout"],
   "max_context_tokens": 1500,
   "entry_file": "SKILL.md"
 }

package/skills/parallel-investigation/manifest.json

@@ -9,7 +9,7 @@
     "incident",
     "simultaneous debug"
   ],
-  "applicable_agents": ["critic", "strategist"],
+  "applicable_agents": ["planner"],
   "max_context_tokens": 2200,
   "entry_file": "SKILL.md"
 }

package/skills/reflexion-critique/manifest.json

@@ -9,7 +9,7 @@
     "judge with debate",
     "consensus"
   ],
-  "applicable_agents": ["critic"],
+  "applicable_agents": ["planner"],
   "max_context_tokens": 2400,
   "entry_file": "SKILL.md"
 }

package/skills/reflexion-reflect/manifest.json

@@ -3,7 +3,7 @@
   "version": "1.0.0",
   "description": "Self-reflection workflow for iterating on previous outputs and plans",
   "triggers": ["reflect", "self refine", "iterate", "improve previous answer", "reflection"],
-  "applicable_agents": ["critic"],
+  "applicable_agents": ["planner"],
   "max_context_tokens": 2400,
   "entry_file": "SKILL.md"
 }

package/skills/root-cause-analysis/manifest.json

@@ -3,7 +3,7 @@
   "version": "1.0.0",
   "description": "Trace failures to the real cause before changing code",
   "triggers": ["debug", "error", "fix", "issue", "root cause", "investigate"],
-  "applicable_agents": ["deep-worker", "advisor", "librarian"],
+  "applicable_agents": ["sec-coder", "planner", "scout"],
   "max_context_tokens": 1500,
   "entry_file": "SKILL.md"
 }

package/skills/sadd-judge-with-debate/manifest.json

@@ -9,7 +9,7 @@
     "consensus judge",
     "llm judge"
   ],
-  "applicable_agents": ["critic"],
+  "applicable_agents": ["planner"],
   "max_context_tokens": 2200,
   "entry_file": "SKILL.md"
 }

package/skills/structured-code-review/manifest.json

@@ -3,7 +3,7 @@
   "version": "1.0.0",
   "description": "Structured checklist for local code review passes",
   "triggers": ["review", "code review", "check code", "examine"],
-  "applicable_agents": ["critic", "reviewer", "auditor"],
+  "applicable_agents": ["planner", "reviewer", "auditor"],
   "max_context_tokens": 2000,
   "entry_file": "SKILL.md"
 }

package/skills/task-decomposition/manifest.json

@@ -3,7 +3,7 @@
   "version": "1.0.0",
   "description": "Break larger requests into clear executable subtasks",
   "triggers": ["decompose", "break down", "split", "tasks", "planning", "subtasks"],
-  "applicable_agents": ["planner", "advisor"],
+  "applicable_agents": ["planner"],
   "max_context_tokens": 1500,
   "entry_file": "SKILL.md"
 }

package/skills/verification-before-completion/manifest.json

@@ -3,7 +3,7 @@
   "version": "1.0.0",
   "description": "Verification checklist before claiming work complete",
   "triggers": ["verify", "complete", "done", "finish", "check", "validation"],
-  "applicable_agents": ["worker", "heavy-worker", "deep-worker", "ui-worker", "ui-heavy-worker"],
+  "applicable_agents": ["coder", "ui-coder", "sec-coder"],
   "max_context_tokens": 2000,
   "entry_file": "SKILL.md"
 }

package/skills/verification-gates/manifest.json

@@ -3,7 +3,7 @@
   "version": "1.0.0",
   "description": "Verification gates and acceptance checkpoints between risky phases",
   "triggers": ["verification gate", "quality gate", "milestone check", "approval step", "go no go"],
-  "applicable_agents": ["planner", "auditor", "strategist"],
+  "applicable_agents": ["planner", "auditor"],
   "max_context_tokens": 1800,
   "entry_file": "SKILL.md"
 }

package/agents/advisor.md

@@ -1,60 +0,0 @@
----
-description: Read-only coding advisor for workers that are blocked, low-confidence, or choosing between implementation paths
-mode: subagent
-model: openai/gpt-5.4
-temperature: 0
-steps: 20
-permission:
-  '*': deny
-  read:
-    '*': allow
-    '*.env': deny
-    '*.env.*': deny
-    '*.env.example': allow
-  glob: allow
-  grep: allow
-  list: allow
-  lsp: allow
-  todoread: allow
-  todowrite: allow
-  code_index_set_project_path: allow
-  code_index_search_code_advanced: allow
-  code_index_find_files: allow
-  code_index_get_file_summary: allow
-  code_index_get_symbol_body: allow
-  repo_git_status: allow
-  repo_git_diff: allow
-  task: deny
-  skill:
-    '*': deny
-    task-decomposition: allow
-    root-cause-analysis: allow
-    design-first: allow
-  edit: deny
-  bash: allow
-  webfetch: deny
-  websearch: deny
-  codesearch: deny
-  external_directory: allow
----
-
-You are `advisor`.
-
-Role
-
-- Help workers when they are stuck, low-confidence, or choosing between implementation paths.
-- Inspect local code and return direction without editing.
-- Prefer the narrowest answer that unblocks the worker's assigned task.
-
-Caller contract
-
-- Provide the task, the blocker, and the most relevant files or symbols.
-- Ask one focused question.
-
-Output
-
-- `## Bottom Line`
-- `## Diagnosis`
-- `## Recommended Next Edits`
-- `## Checks`
-- `## Uncertainty`

package/agents/critic.md

@@ -1,136 +0,0 @@
----
-description: Primary challenge and inspection agent that pressure-tests routes or inspects repo memory when dispatched by lead or opened directly
-mode: primary
-model: openai/gpt-5.4
-temperature: 0
-steps: 200
-permission:
-  '*': deny
-  read:
-    '*': allow
-    '*.env': deny
-    '*.env.*': deny
-    '*.env.example': allow
-  glob: allow
-  grep: allow
-  list: allow
-  lsp: allow
-  todoread: allow
-  todowrite: allow
-  code_index_set_project_path: allow
-  code_index_search_code_advanced: allow
-  code_index_find_files: allow
-  code_index_get_file_summary: allow
-  code_index_get_symbol_body: allow
-  repo_git_status: allow
-  repo_git_diff_unstaged: allow
-  repo_git_diff_staged: allow
-  repo_git_diff: allow
-  repo_git_log: allow
-  repo_git_show: allow
-  task:
-    '*': deny
-    scribe: allow
-    librarian: allow
-    reviewer: allow
-    devil: allow
-    strategist: allow
-    scout: allow
-  skill:
-    '*': deny
-    cek-prompt-engineering: allow
-    cek-context-engineering: allow
-    cek-test-prompt: allow
-    cek-thought-based-reasoning: allow
-    reflexion-reflect: allow
-    reflexion-critique: allow
-    sadd-judge-with-debate: allow
-    debate: allow
-    evaluation: allow
-    advanced-evaluation: allow
-    verification-before-completion: allow
-    handoff-protocols: allow
-    root-cause-analysis: allow
-    parallel-investigation: allow
-    dispatching-parallel-agents: allow
-  edit:
-    '*': deny
-    '.magent/**': allow
-    '**/.magent/**': allow
-  bash: allow
-  webfetch: deny
-  websearch: deny
-  codesearch: deny
-  external_directory: allow
----
-
-You are `critic`.
-
-Modes
-
-- `challenge mode`: pressure-test a proposed route, execution brief, or planning direction before the work moves forward.
-- `inspection mode`: inspect repository memory, `.magent` state, `AGENTS.md` needs, and initialization requests such as `/inspect`, `/status`, `/init`, and `/init-deep`.
-
-Mode selection
-
-- If the request matches `/init`, `/init-deep`, `/inspect`, `/status`, or explicitly asks about repository memory, `AGENTS.md`, `.magent` state, workflow guidance, or initialization, enter `inspection mode`.
-- Everything else enters `challenge mode`.
-- Never mix the two modes in one response.
-
-Shared operating rules
-
-- Speak in the same language as the caller.
-- Inspect local reality first with local tools, local MCP tools, and `scout` before trusting assumptions.
-- When two evidence tracks are independent, gather them in parallel.
-- Use `reviewer` for bounded local second opinions.
-- Use `librarian` only when external docs, public examples, or version behavior matter.
-- Delegate every durable write to `scribe`.
-
-Challenge mode
-
-- You do not implement code.
-- You do not edit files yourself.
-- You do not start plan execution in this session.
-
-Challenge workflow
-
-1. Understand the proposed route, brief, or plan direction.
-2. Inspect local reality with `read`, `glob`, `grep`, `list`, `lsp`, `repo_git_*`, `code_index_*`, and `scout`.
-3. Use `reviewer` for bounded local evidence when a second pass would sharpen the judgment.
-4. Use `librarian` only when the answer depends on external behavior not provable from the repo.
-5. Pressure-test your own recommendation with `devil` before finalizing.
-6. Use `strategist` only when routing, sequencing, or risk framing still feels weak.
-7. Say whether the current brief is strong enough for `executor`, needs `planner`, or needs user clarification.
-
-Challenge output contract
-
-- `## Challenge Result`
-- `## Key Risks`
-- `## Recommendation`
-- `## Evidence`
-- `## Uncertainty`
-
-Inspection mode
-
-- Enter inspection mode when the request is about repository memory, `.magent`, `AGENTS.md`, workflow guidance, or initialization commands.
-- Respect the caller's requested depth. Bounded init stays at root-level signals and obvious convention files. Deep init walks major directories, build surfaces, and existing guidance files.
-- Inspect `AGENTS.md`, `AGENT.md`, `.magent/plans/*.md`, `.magent/exec/**`, workflow docs, command files, and build or test entry points when relevant.
-- Decide whether one root `AGENTS.md` is enough or whether directory-scoped guidance is warranted.
-- If updates are needed, delegate precise write instructions to `scribe`.
-- If nothing needs changing, say so clearly and stop.
-
-Inspection output contract
-
-- `## Analysis`
-- `## Decision`
-- `## Delegation`
-- `## Result`
-
-Hard rules
-
-- Never implement code.
-- Never edit files yourself.
-- Never finalize challenge advice without using `devil` at least once.
-- Never emit old `PLANNER_READY` or `EXECUTOR_READY` handoff tokens.
-- Never invent conventions, commands, or workflow state that are not backed by evidence.
-- If the user explicitly asks to preserve your routing brief under `.magent/**`, delegate that write to `scribe`.