Team UI Polish

Automatic UI quality improvement pipeline: scan -> diagnose -> optimize -> verify. Built on team-worker agent architecture -- all worker roles share a single agent definition with role-specific Phase 2-4 loaded from roles/<role>/role.md.

Architecture

Skill(skill="team-ui-polish", args="task description")
                    |
         SKILL.md (this file) = Router
                    |
     +--------------+--------------+
     |                             |
  no --role flag              --role <name>
     |                             |
  Coordinator                  Worker
  roles/coordinator/role.md    roles/<name>/role.md
     |
     +-- analyze -> dispatch -> spawn workers -> STOP
                                    |
                    +-------+-------+-------+
                    v       v       v       v
           [team-worker agents, each loads roles/<role>/role.md]
          scanner  diagnostician  optimizer <-> verifier (GC loop)

Role Registry

Role Router

Parse $ARGUMENTS:

• Has --role <name> -> Read roles/<name>/role.md, execute Phase 2-4
• No --role -> roles/coordinator/role.md, execute entry router

Delegation Lock

Coordinator is a PURE ORCHESTRATOR. It coordinates, it does NOT do.

Before calling ANY tool, apply this check:

If a tool call is BLOCKED: STOP. Create a task, spawn a worker.

No exceptions for "simple" tasks. Even a single-file read-and-report MUST go through spawn_agent.

Shared Constants

• Session prefix: UIP
• Session path: .workflow/.team/UIP-<date>-<slug>/
• CLI tools: ccw cli --mode analysis (read-only), ccw cli --mode write (modifications)
• Message bus: mcp__ccw-tools__team_msg(session_id=<session-id>, ...)
• Max GC rounds: 2

Worker Spawn Template

Coordinator spawns workers using this template:

spawn_agent({
  agent_type: "team_worker",
  task_name: "<task-id>",
  fork_turns: "none",
  message: `## Role Assignment
role: <role>
role_spec: <skill_root>/roles/<role>/role.md
session: <session-folder>
session_id: <session-id>
requirement: <task-description>
inner_loop: <true|false>

Read role_spec file (<skill_root>/roles/<role>/role.md) to load Phase 2-4 domain instructions.

## Task Context
task_id: <task-id>
title: <task-title>
description: <task-description>
pipeline_phase: <pipeline-phase>

## Upstream Context
<prev_context>`
})

After spawning, use wait_agent({ timeout_ms: 1800000 }) to collect results. If result.timed_out, send STATUS_CHECK via followup_task (wait 3 min), then FINALIZE with interrupt (wait 3 min), then mark timed_out and close agents. Use close_agent({ target }) each worker.

Model Selection Guide

UI polish is a sequential pipeline where scan findings drive all downstream work. Scanner and verifier need thorough analysis, optimizer needs creative problem-solving.

Scan-to-Diagnosis Context Flow

Scanner findings must reach diagnostician via coordinator's upstream context:

// After SCAN-001 completes, coordinator sends findings to diagnostician
spawn_agent({
  agent_type: "team_worker",
  task_name: "DIAG-001",
  fork_turns: "none",
  message: `## Upstream Context
Scan report: <session>/scan/scan-report.md`
})

User Commands

Specs Reference

• specs/pipelines.md -- Pipeline definitions and task registry
• specs/team-config.json -- Team configuration
• specs/anti-patterns.md -- AI slop and design anti-pattern catalog
• specs/design-standards.md -- Impeccable positive design standards
• specs/fix-strategies.md -- Issue-to-fix mapping
• specs/scoring-guide.md -- Scoring rubric

Session Directory

.workflow/.team/UIP-<date>-<slug>/
+-- .msg/
|   +-- messages.jsonl         # Team message bus
|   +-- meta.json              # Pipeline config + GC state
+-- scan/                      # Scanner output
|   +-- scan-report.md
+-- diagnosis/                 # Diagnostician output
|   +-- diagnosis-report.md
+-- optimization/              # Optimizer output
|   +-- fix-log.md
+-- verification/              # Verifier output
|   +-- verify-report.md
+-- evidence/                  # Screenshots (before/after)
|   +-- *.png
+-- wisdom/                    # Cross-task knowledge

v4 Agent Coordination

Message Semantics

Agent Health Check

Use list_agents({}) in handleResume and handleComplete:

// Reconcile session state with actual running agents
const running = list_agents({})
// Compare with tasks.json active tasks
// Reset orphaned tasks (in_progress but agent gone) to pending

Named Agent Targeting

Workers are spawned with task_name: "<task-id>" enabling direct addressing:

• send_message({ target: "DIAG-001", message: "..." }) -- send additional scan findings to diagnostician
• followup_task({ target: "OPT-001", message: "..." }) -- assign optimization from diagnosis report
• close_agent({ target: "VERIFY-001" }) -- cleanup after verification

Error Handling