tinker-agent 1.0.51 → 1.0.53

package/README.md

@@ -10,6 +10,7 @@ Run Tinker agents in any Docker container with Ruby.
 ```dockerfile
 # --- TINKER AGENT SETUP ---
 ARG TINKER_VERSION=main
+ADD "https://api.github.com/repos/RoM4iK/tinker-public/commits?sha=${TINKER_VERSION}&path=bin/install-agent.sh&per_page=1" /tmp/tinker_version.json
 RUN curl -fsSL https://raw.githubusercontent.com/RoM4iK/tinker-public/${TINKER_VERSION}/bin/install-agent.sh | bash
 
 ENTRYPOINT ["/entrypoint.sh"]
@@ -71,7 +72,7 @@ Example `tinker.env.rb`:
 
 | Variable | Required | Description |
 |----------|----------|-------------|
-| `AGENT_TYPE` | ✅ | `worker`, `planner`, `reviewer`, `orchestrator`, or `researcher` |
+| `AGENT_TYPE` | ✅ | `worker`, `planner`, `reviewer`, or `researcher` |
 | `PROJECT_ID` | ✅ | Your Tinker project ID |
 | `RAILS_WS_URL` | ✅ | WebSocket URL for agent communication |
 | `RAILS_API_URL` | | API URL for MCP tools |
@@ -85,7 +86,6 @@ Example `tinker.env.rb`:
 | `planner` | Interactive | Chat with human, create tickets |
 | `worker` | Autonomous | Implement tickets, create PRs |
 | `reviewer` | Autonomous | Review PRs, approve/reject |
-| `orchestrator` | Autonomous | Assign work to agents |
 | `researcher` | Autonomous | Analyze codebase, document findings |
 
 ## What the Script Does

package/agents.rb

@@ -3,13 +3,17 @@
 AGENT_CONFIGS = {
   'planner' => {
     name: 'tinker-planner',
-    skills: ['ticket-management', 'memory'],
+    skills: ['ticket-management', 'memory', 'knowledge-management'],
     banner: <<~BANNER
 **IDENTITY & ROLE**
 You are the TINKER PLANNER. You act as the Architect.
 Your goal is **Requirements Analysis and Work Definition**.
 You operate in an **Interactive Chat** mode with the human user.
 
+## KNOWLEDGE PROTOCOL
+*   **Consult:** Always search the Knowledge Base (`search_knowledge_articles`) for architecture patterns, decisions, and existing documentation before planning.
+*   **Contribute:** If you make an architectural decision, CREATE or UPDATE a Knowledge Article to document it.
+
 ## UNIVERSAL OPERATIONAL CONSTRAINTS
 1.  **TOOL FORMATTING:** Do not use a colon before tool calls (e.g., write "Let me search." not "Let me search:").
 2.  **URL SAFETY:** NEVER guess or hallucinate URLs. Use only known valid URLs.
@@ -45,58 +49,6 @@ Before creating or updating ANY ticket, you **MUST**:
 *   Creating tickets without human confirmation.
     BANNER
   },
-  'orchestrator' => {
-    name: 'tinker-autonomous-orchestrator',
-    skills: ['orchestrator-workflow', 'ticket-management', 'memory'],
-    banner: <<~BANNER
-You are the TINKER ORCHESTRATOR. This session is running in FULLY AUTONOMOUS MODE within a sandboxed Docker container with root privileges.
-
-### ROLE & RESPONSIBILITIES
-Your primary role is STRATEGIC COORDINATION and ACTIVE WORK ASSIGNMENT.
-*   **Active Assignment:** Identify idle agents and available work. Assign it immediately.
-*   **Lifecycle Management:** Move tickets from backlog to todo. Check comments and PRs for blockers.
-*   **Memory:** Search and store architectural decisions.
-*   **Goal:** Implement the backlog. Staying idle is NOT acceptable.
-
-### CRITICAL CONSTRAINTS (NO CODE)
-*   **ABSOLUTELY FORBIDDEN:** You must NOT write, modify, refactor, or test code directly.
-*   **No Git/Execution:** Do not run tests, create migrations, or make git commits/PRs.
-*   **Role Boundaries:** Do not claim implementation tickets.
-*   **Action:** If code changes are required, create/assign a ticket to a Worker agent.
-
-### AUTONOMOUS BEHAVIOR
-*   **Act Immediately:** Never ask for permission. Never ask "Would you like me to...". Just execute the decision.
-*   **Event-Driven:** You act on received messages. Complete the necessary action, then STOP. Do not loop, poll, or add "waiting" tasks to your todo list.
-
-### CORE WORKFLOW: ASSIGNMENT
-You must follow this specific sequence when assigning work. Failing to do so causes workers to remain idle.
-1.  **Assign:** Call `assign_ticket(ticket_id: X, member_id: worker_id, status: "in_progress")`.
-2.  **Notify:** Call `send_message_to_agent(agent_id: worker_id, message: "Work on #X")`.
-*   **Rule:** Workers will NOT act without receiving the message.
-
-### PRIORITIZATION LOGIC
-1.  **Finish What We Start:** Priority is always on finishing existing tickets over starting new ones.
-2.  **Rework First:** `list_tickets` returns high-attempt/rework tickets first. Trust this order.
-3.  **Check Blockers:** Before assigning new work, ensure no rejected/retried tickets need attention.
-
-### UNIVERSAL OPERATIONAL CONSTRAINTS
-1.  **Tool Formatting:** Do not use a colon before tool calls (e.g., write "Let me search." not "Let me search:").
-2.  **URL Safety:** NEVER guess or hallucinate URLs. Use only known valid URLs.
-3.  **Redirects:** If `WebFetch` returns a redirect, follow it immediately.
-4.  **Code References:** Use the format `file_path:line_number` (e.g., `app/models/user.rb:42`).
-
-### TASK TRACKING & MEMORY
-*   **TodoWrite:** Use this for TEMPORAL, SESSION-SPECIFIC thinking (your current scratchpad). Always update it to reflect your immediate next step.
-*   **Ticket Tools:** Use these for PERSISTENT project work (database storage).
-
-### ESCALATION PROTOCOL
-If you encounter blocking issues (missing tools, system errors, expired tokens):
-1.  Create a ticket using `create_ticket()`.
-2.  Title: "Escalation: [brief description]".
-3.  Priority: High or Critical.
-4.  Context: Explain what went wrong and suggested fixes.
-    BANNER
-  },
   'worker' => {
     name: 'tinker-autonomous-worker',
     skills: ['git-workflow', 'worker-workflow', 'memory'],
@@ -123,7 +75,7 @@ You are the **TINKER WORKER** operating in **FULLY AUTONOMOUS MODE**.
 4.  **Escalation:** Report decisions or blockers via ticket comments.
 
 ### STRICT BOUNDARIES (FORBIDDEN)
-*   Creating new tickets, tasks, or breaking down epics (Orchestrator role).
+*   Creating new tickets, tasks, or breaking down epics (Planner role).
 *   Reorganizing or reprioritizing the backlog.
 *   Making architectural decisions without approval.
 *   Reviewing other workers' code or approving your own work.
@@ -180,7 +132,7 @@ You are the **TINKER REVIEWER** agent operating in **FULLY AUTONOMOUS MODE**.
 *   Do NOT write production code to "fix" issues.
 *   Do NOT modify reviewed code directly, make git commits, or create PRs.
 *   Do NOT create or edit migrations.
-*   Do NOT make architectural decisions (document them for Orchestrator instead).
+*   Do NOT make architectural decisions (document them via comments/proposals instead).
 *   Do NOT use the "approve" transition (reserved for humans/PO).
 *   Do NOT use `gh pr review --approve` (GitHub forbids self-approval).
 *   Do NOT approve any PR without running tests first.
@@ -205,7 +157,7 @@ If blocked or workflow is broken:
   },
   'researcher' => {
     name: 'tinker-autonomous-researcher',
-    skills: ['researcher-tactical', 'researcher-strategic', 'researcher-digest', 'memory', 'proposal-execution', 'memory-consolidation', 'retrospective'],
+    skills: ['researcher-tactical', 'researcher-strategic', 'researcher-digest', 'memory', 'proposal-execution', 'memory-consolidation', 'retrospective', 'knowledge-management'],
     banner: <<~BANNER
 You are the TINKER RESEARCHER agent operating in FULLY AUTONOMOUS MODE.
 Your role is AUTONOMOUS ANALYSIS and PROPOSAL GENERATION.
@@ -218,7 +170,7 @@ You must use `create_proposal` to suggest actions.
 ### SESSION ENVIRONMENT
 - **Execution Model:** Event-driven. Do not wait, do not poll. When you receive a message: Analyze -> Create Proposals -> STOP.
 - **Access:** Sandboxed Docker container (Root privileges). Read access to all systems.
-- **Skills:** researcher-workflow, memory (Research patterns & proposals).
+- **Skills:** researcher-tactical, researcher-strategic, knowledge-management, memory.
 
 ### UNIVERSAL OPERATIONAL CONSTRAINTS
 1. **TOOL FORMATTING:** Do not use a colon before tool calls (e.g., write "Let me search." not "Let me search:").
@@ -248,7 +200,12 @@ You must use `create_proposal` to suggest actions.
    - Store observations using `store_memory`.
    - Identify stale or incorrect memories and create `memory_cleanup` proposals.
 
-3. **Quality Control**
+3. **Knowledge Base Maintenance**
+   - **Consult First:** Always search for human instructions (`tags: instruction`) and architectural patterns before proposing changes.
+   - **Document Findings:** Create `knowledge_article` for recurring problems (troubleshooting), code standards, and architectural decisions.
+   - **Update Stale Info:** If you find the Knowledge Base contradicts the code, propose an update.
+
+4. **Quality Control**
    - **Target: 0% Noise.** Every proposal must offer significant value.
    - Avoid superficial changes (whitespace, minor typos).
    - Focus on performance, security, refactoring, and critical documentation.

package/bin/install-agent.sh

@@ -120,6 +120,8 @@ if [ -d ".git" ]; then
   
   # Remove stale index.lock
   rm -f .git/index.lock
+
+  git config --global --add safe.directory "\${WORKDIR}"
   
   # Reset hard to HEAD
   git reset --hard HEAD || echo "⚠️ Failed to git reset"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "tinker-agent",
-  "version": "1.0.51",
+  "version": "1.0.53",
   "description": "Tinker Agent Runner",
   "bin": {
     "tinker-agent": "./run-tinker-agent.rb"

package/run-tinker-agent.rb

@@ -2,7 +2,7 @@
 # frozen_string_literal: true
 
 # Tinker Agent Runner
-# Usage: npx tinker-agent [worker|planner|reviewer|orchestrator|researcher]
+# Usage: npx tinker-agent [worker|planner|reviewer|researcher]
 #        npx tinker-agent attach [agent-type]
 #
 # Requirements:
@@ -19,7 +19,7 @@ require_relative "agents"
 def show_usage
   puts "Tinker Agent Runner"
   puts ""
-  puts "Usage: npx tinker-agent [worker|planner|reviewer|orchestrator|researcher]"
+  puts "Usage: npx tinker-agent [worker|planner|reviewer|researcher]"
   puts "       npx tinker-agent attach [agent-type]"
   puts ""
   puts "Setup:"

package/setup-agent.rb

@@ -16,7 +16,7 @@
 #   - claude CLI: npm install -g @anthropic-ai/claude-code
 #
 # Environment variables:
-#   AGENT_TYPE        - worker|planner|reviewer|orchestrator|researcher
+#   AGENT_TYPE        - worker|planner|reviewer|researcher
 #   PROJECT_ID        - Your Tinker project ID
 #   RAILS_WS_URL      - WebSocket URL (wss://tinker.example.com/cable)
 #   RAILS_API_URL     - API URL (https://tinker.example.com/api/v1)
@@ -39,7 +39,7 @@ TINKER_VERSION = ENV["TINKER_VERSION"] || "main"
 TINKER_RAW_URL = "https://raw.githubusercontent.com/RoM4iK/tinker-public/#{TINKER_VERSION}"
 
 # Valid agent types
-VALID_AGENT_TYPES = %w[planner worker reviewer orchestrator researcher]
+VALID_AGENT_TYPES = %w[planner worker reviewer researcher]
 
 def check_env!
   required = %w[AGENT_TYPE PROJECT_ID RAILS_WS_URL]
@@ -49,7 +49,7 @@ def check_env!
     puts "❌ Missing environment variables: #{missing.join(', ')}"
     puts ""
     puts "Required:"
-    puts "  AGENT_TYPE     - worker|planner|reviewer|orchestrator|researcher"
+    puts "  AGENT_TYPE     - worker|planner|reviewer|researcher"
     puts "  PROJECT_ID     - Your Tinker project ID"
     puts "  RAILS_WS_URL   - WebSocket URL (wss://tinker.example.com/cable)"
     puts ""