capital-allocator

name	capital-allocator
description	Makes capital allocation decisions based on market regime and agent performance. Adjusts governance policy to control entry modes, kelly fractions, and per-agent allocation limits.
version	1.0.0
user-invocable	true
metadata	{"openclaw":{"requires":{"bins":["curl","jq"]}}}

Capital Allocator

Make capital allocation decisions by combining market regime analysis with per-agent performance metrics, then execute adjustments via governance policy updates.

Allocation Policy Table

Regime	Crypto Entry Mode	Kelly Fraction	Max Intent % of Capital
HighVol	vol_straddle	0.15	50%
LowVol	arb_only	0.30	100%
Trending	directional	0.25	100%
Ranging	arb_only	0.20	75%

Step 1: Gather Current State

Get risk state and positions

RISK=$(curl -sf $PLOY_API_BASE/api/sidecar/risk -H "x-ploy-sidecar-token: $PLOY_SIDECAR_AUTH_TOKEN")
echo "$RISK" | jq '{risk_state: .risk_state, daily_pnl: .daily_pnl_usd, drawdown: .current_drawdown_usd, positions: (.positions | length), queue: .queue_depth}'

Get current governance policy

POLICY=$(curl -sf $PLOY_API_BASE/api/governance/policy -H "x-ploy-admin-token: $PLOY_API_ADMIN_TOKEN")
echo "$POLICY" | jq '{block_new_intents: .block_new_intents, blocked_domains: .blocked_domains, max_intent: .max_intent_notional_usd, max_total: .max_total_notional_usd, metadata: .metadata}'

Get governance status (includes agent states)

STATUS=$(curl -sf $PLOY_API_BASE/api/governance/status -H "x-ploy-admin-token: $PLOY_API_ADMIN_TOKEN")
echo "$STATUS" | jq .

Get system capabilities (active agents/domains)

CAPS=$(curl -sf $PLOY_API_BASE/api/capabilities)
echo "$CAPS" | jq '{active_domains: .active_domains, coordinator: .coordinator_running, deployments: .scoped_enabled_deployments}'

Get PnL history for performance scoring

PNL=$(curl -sf $PLOY_API_BASE/api/stats/pnl)
echo "$PNL" | jq '.'

Step 2: Detect Current Regime

Use the regime-detector skill OR read the last confirmed regime from memory.

If no regime is recorded, run the regime-detector skill first.

Step 3: Score Agent Performance

For each active agent/domain, compute a performance score:

Composite Score Formula:

score = 0.4 * sharpe_normalized + 0.3 * win_rate + 0.3 * (1 - drawdown_ratio)

Where:

• sharpe_normalized = clamp(sharpe / 2.0, 0, 1) — Sharpe of 2.0+ gets perfect score
• win_rate = wins / total_trades (from PnL data)
• drawdown_ratio = current_drawdown / max_allowed_drawdown

Score interpretation:

• > 0.70: Excellent — increase allocation
• 0.40 - 0.70: Normal — maintain allocation
• < 0.40: Underperforming — consider pausing

Step 4: Make Allocation Decisions

For each agent:

1. If score < 0.30 AND not paused AND more than 1 agent active:

   • Pause the agent via domain control
   • Record pause timestamp in memory for cooldown tracking

2. If paused AND score >= 0.50 AND paused > 15 minutes ago:

   • Resume the agent via domain control

3. Adjust per-agent allocation:

   • agent_max_alloc = score * regime_max_intent_pct
   • Clamp to max 40% of total capital per single agent

4. SAFETY GUARD: Never pause ALL agents. If pausing would leave 0 active agents, skip.

Step 5: Execute Decisions

Pause a domain

curl -sf -X POST $PLOY_API_BASE/api/system/pause \
  -H "Content-Type: application/json" \
  -H "x-ploy-admin-token: $PLOY_API_ADMIN_TOKEN" \
  -d '{"domain": "crypto"}' | jq .

Resume a domain

curl -sf -X POST $PLOY_API_BASE/api/system/resume \
  -H "Content-Type: application/json" \
  -H "x-ploy-admin-token: $PLOY_API_ADMIN_TOKEN" \
  -d '{"domain": "crypto"}' | jq .

Update governance with new allocation

curl -sf -X PUT $PLOY_API_BASE/api/governance/policy \
  -H "Content-Type: application/json" \
  -H "x-ploy-admin-token: $PLOY_API_ADMIN_TOKEN" \
  -d '{
    "block_new_intents": false,
    "blocked_domains": [],
    "max_intent_notional_usd": <COMPUTED>,
    "max_total_notional_usd": <COMPUTED>,
    "updated_by": "openclaw.capital-allocator",
    "reason": "Allocation update: regime=<REGIME>, scores={crypto: X.XX, sports: Y.YY}"
  }' | jq .

Step 6: Record & Report

Write allocation decision to memory:

allocation_decision:
  timestamp: <ISO8601>
  regime: <regime>
  scores: {crypto: X.XX, sports: Y.YY, politics: Z.ZZ}
  actions: [pause crypto | resume sports | adjust limits]
  new_max_intent: $XX.XX
  new_max_total: $XXX.XX

Report format:

Capital Allocation Update
─────────────────────────
Regime: <REGIME>
| Agent    | Score | Status  | Max Alloc |
|----------|-------|---------|-----------|
| crypto   | 0.72  | active  | 25%       |
| sports   | 0.45  | active  | 15%       |
| politics | 0.28  | PAUSED  | 0%        |
─────────────────────────
Action: Paused politics (score 0.28 < 0.30 threshold)
New limits: intent=$75, total=$300