DQIII8: Skill for Claude Code

/test-team — Agent Team Coordination Test

Direct coordination test between agents using Agent Teams. Demonstrates that the output of one agent feeds directly into the next.

Team

Task: Implement Kelly Criterion in Python based on prior research.

Agent 1 — research-analyst (first)

Researches the Kelly Criterion and produces a structured summary with:

• Exact mathematical formula: f* = (bp - q) / b where b = net odds, p = win probability, q = 1 - p
• Input parameters and their valid ranges
• Half-Kelly variant (f* / 2) and when to prefer it
• Use cases in systematic trading (position sizing)
• Known limitations (sensitivity to p estimation)

Writes result to: tasks/results/research-kelly-[timestamp].md

Agent 2 — python-specialist (after Agent 1)

Reads the research-analyst result from tasks/results/research-kelly-*.md and based on it implements:

def kelly_criterion(win_prob: float, win_loss_ratio: float, half_kelly: bool = True) -> float:
    """
    Calculates the optimal position size according to Kelly Criterion.
    ...
    """

Implementation requirements:

• Full type hints
• Input validation (0 < win_prob < 1, win_loss_ratio > 0)
• Half-Kelly support (default True — more conservative)
• Usage example in docstring with real trading values

Writes result to: tasks/results/python-kelly-[timestamp].md

Coordination protocol

research-analyst → tasks/results/research-kelly-*.md
                              ↓
python-specialist reads that file → implements function

The python-specialist does NOT start until research-analyst has written its result. This validates sequential coordination of Agent Teams.

Execution

Launch both agents as a coordinated team. The orchestrator waits for the result from research-analyst before passing context to python-specialist. When done, show:

1. Research summary (formula + parameters)
2. Implemented Python code
3. Confirmation: [TEAM] Kelly Criterion — research + impl complete