Agent skill

techartist-workflow

Tech Artist orchestration - startup sequence, workflow execution, message handling, exit conditions. Use when starting Tech Artist tasks.

Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/data/techartist-workflow

SKILL.md

Tech Artist Orchestration

"The conductor of visual creation - coordinates workflow from assignment to delivery."


Startup Workflow

  1. Load Skill("ta-router") to understand your skill set

  2. Load Skill("threejs-builder") to understand your Three.js skill set

  3. Load Skill("shared-validation-feedback-loops") to understand validation loops

  4. Process pending messages - IMPORTANT: Messages are in the master branch, accessed via relative path like ../agentic-threejs/.claude/session/messages/techartist/. Consolidate all the .json messages requests and delete the files before continue. Update watchdog status.

  5. Read current-task-techartist.json - IMPORTANT: State file is in the master branch, accessed via relative path like ../agentic-threejs/.claude/session/current-task-techartist.json. Reason about the message request and define your next action


Dashboard Status Update - CRITICAL: Before starting ANY work action, update your status in current-task-techartist.json:

Action Update State File Like This Send Status Update to Watchdog
Starting work on task state.status = "working" + id = "{taskId}" + state.lastSeen = "{ISO_TIMESTAMP}" Send-StatusUpdate -From "techartist" -Status "working" -CurrentTask "{taskId}"
Blocked by question state.status = "awaiting_pm" + state.lastSeen = "{ISO_TIMESTAMP}" Send-StatusUpdate -From "techartist" -Status "waiting"
Sending to QA state.status = "idle" + id = null + state.lastSeen = "{ISO_TIMESTAMP}" Send-StatusUpdate -From "techartist" -Status "idle"
Self-reporting progress state.lastSeen = "{ISO_TIMESTAMP}" -

Implementation Workflow

PRE-REQUISITE: You should already be in your worktree directory (../techartist-worktree) before starting this workflow!

  1. UPDATE STATE FILE (MANDATORY - First step)

    • Edit current-task-techartist.json:
      • state.status = "working"
      • state.currentTaskId = "{taskId}"
      • state.lastSeen = "{ISO_TIMESTAMP}"
    • Send status update to watchdog: Send-StatusUpdate -From "techartist" -Status "working" -CurrentTask "{taskId}"
  2. Task Research (MANDATORY Before Coding new tasks)

    • Step 1: GDD Reading (Bugs can skip it)

      Skill("dev-research-gdd-reading")
      
      • Read docs/design/gdd/index.md for overview
      • Read feature-specific GDD files
      • Check decision log and open questions
    • Step 2: Codebase Exploration

      Task({
      subagent_type: "developer-code-research",
      description: "Research patterns for {feature}",
      prompt: "Research existing codebase patterns for implementing {feature}",
      timeout: 300000
      })
      
  3. IMPLEMENTATION

    • Create/modify files following researched patterns
  4. TEST COVERAGE CHECK (MANDATORY - CANNOT SKIP)

    • Skill("qa-test-creation") - Check if tests exist for modified files
    • If tests missing: MUST invoke test-creator sub-agent before proceeding
    • Files changed without tests = BLOCKING - cannot proceed to step 8

    ⚠️ NON-BYPASSABLE GATE:

    • NO exceptions for "bug fixes", "refactorings", or "non-visual changes"
    • Even trivial changes need test coverage (unit test minimum)
    • ONLY PM can approve skipping tests via explicit message
  5. IF BLOCKED

    • Update state: state.status = "awaiting_pm"
    • Send question to PM using the Write tool:
    Write to: .claude/session/messages/pm/msg-pm-{timestamp}-001.json
    Content:
    {
      "id": "msg-pm-{timestamp}-001",
      "from": "techartist",
      "to": "pm",
      "type": "question",
      "priority": "high",
      "payload": {
        "question": "How should I handle X?",
        "context": "Current situation..."
      },
      "timestamp": "{ISO-8601-timestamp}",
      "status": "pending"
    }
    
    • Document blocker in task memory
    • Exit and wait
  6. FEEDBACK LOOPS

  7. COMMIT - At the end of the task, commit all changes to your branch

  8. SEND TO QA

  • Update state: state.status = "idle", id = null

  • Send status update to watchdog: Send-StatusUpdate -From "techartist" -Status "idle"

  • Send completion using the Write tool:

    Write to: .claude/session/messages/pm/msg-pm-{timestamp}-001.json
    Content:
    {
      "id": "msg-pm-{timestamp}-001",
      "from": "techartist",
      "to": "pm",
      "type": "task_complete",
      "priority": "normal",
      "payload": {
        "taskId": "{taskId}",
        "success": true,
        "summary": "Implementation complete"
      },
      "timestamp": "{ISO-8601-timestamp}",
      "status": "pending"
    }
    
  1. EXIT

Expand your agent's capabilities with these related and highly-rated skills.

Didn't find tool you were looking for?

Be as detailed as possible for better results