mcpField
C
cobra-strategy

cobra-strategy

github.com/senpi-aiOther14043 installsUpdated today
Repository

About cobra-strategy

COBRA is a Senpi trading strategy skill focused on short-term Arena performance through concentrated trades in the top SM asset. It combines gated entry signals, fee-optimized limit orders, and DSL-based exit management.

Key features

  1. 01Trades only the top SM asset
  2. 02Scores dominant-asset entry signals
  3. 03Uses fee-optimized limit orders
  4. 04DSL configuration for exits and timeouts
  5. 05Limits entries and concurrent positions

Use cases

  1. 01Configure a concentrated Arena trading strategy
  2. 02Screen the leading SM asset for entry signals
  3. 03Set position exits and timeouts through DSL

MCP support

The skill instructs the cron to call create_position via Senpi MCP when the scanner produces an entry signal.

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/senpi-ai/senpi-skills --skill cobra-strategy

After installing, type/cobra-strategyin Claude Code to confirm it loaded.

Bundled files

References and scripts installed along with SKILL.md.

README.md
runtime.yaml
config/
cobra-config.json
references/
skill-attribution.md
scripts/
cobra-scanner.py
cobra_config.py

Details

From the project README

🐍 COBRA v1.1 — Arena Sprint Predator

SM dominates. Cobra strikes. One asset. Maximum conviction.

What Makes Cobra Different

Every other Predator agent is built for long-term edge — patient entry, perfect scores, wide DSL. Cobra is built to win Arena weeks.

The Arena rewards ROE% over 7 days. The winning strategy is 2-4 massive conviction trades that ride the dominant SM trend. The #1 Arena agent won with +62% ROE on just 12 trades.

Fee Lesson from the Fleet

The Predators data tells a clear story about fees:

AgentTradesFeesProfitProfit:Fee Ratio
Condor v2.05$2.71+$37.4313.8:1
Cheetah53$23.12+$40.581.8:1
Polar170$239.27+$38.650.16:1

Cobra targets Condor's efficiency: few trades, big impact, minimal fees.

Cobra's approach:
  • Only trades the #1 SM asset — never scatters across multiple assets
  • $400 margin per trade (40% of budget) for concentrated ROE impact
  • FEE_OPTIMIZED_LIMIT orders — maker first, taker fallback at 30s. ~50% fee savings vs market orders.
  • Wide DSL — 180min timeout, 8% retrace, 45min dead weight. Let positions breathe.
  • 3 entries per day max — quality over quantity
  • 10x leverage on majors (BTC/ETH/SOL/HYPE), 5x on others

⛔ CRITICAL AGENT RULES

RULE 1: Install path is /data/workspace/skills/cobra-strategy/
RULE 2: THE SCANNER DOES NOT EXIT POSITIONS — DSL only.
RULE 3: MAX 1 POSITION at a time.
RULE 4: MAX 3 ENTRIES PER DAY. 90-minute cooldown. 120-minute per-asset cooldown.
RULE 5: ONLY trade the #1 SM dominant asset. Never scatter.
RULE 6: XYZ equities are BANNED at parse level.
RULE 7: USE FEE_OPTIMIZED_LIMIT for all entries. Never market orders.

Scanner Logic (2 API calls)

1. leaderboard_get_markets (limit=20)
   → Find #1 non-XYZ asset by SM concentration
   → Gate: Must be >10% SM dominance

2. Score the dominant asset:
   - SM Dominance: >15% = +3, >10% = +2, >5% = +1
   - Contribution Surge: >5% = +2, >2% = +1
   - 4H Trend Confirms: >0.5% in direction = +1
   - 1H Trend Confirms: >0.2% in direction = +1
   - Deep Consensus: >200 traders = +1

   MIN_SCORE: 5/8
   
   → If score >= 5: ENTER with $400 margin at 10x (majors) or 5x
   → Order type: FEE_OPTIMIZED_LIMIT (maker first, taker fallback 30s)

DSL Configuration (Fee-Conscious, Position-Breathing)

Phase 1 (Loss Protection):

  • Max loss: -15%
  • Retrace from high water: 8% (wide — let volatile assets breathe)
  • 3 consecutive breaches required
  • Phase 1 max time: 45 minutes

Phase 2 (Profit Locking — let winners run):

  • +5% → lock 25% of high water
  • +10% → lock 45%
  • +15% → lock 60%
  • +20% → lock 75%
  • +30% → lock 85%
  • +50% → lock 92%

Timeouts:

  • Hard timeout: 180 minutes (3 hours — positions need time)
  • Weak peak cut: 60 minutes at +3% minimum
  • Dead weight cut: 45 minutes

Why This Wins Arena

  1. Concentration: $400 on one trade. A 10% move at 10x = +$400 = 40% ROE on the $1K account.

  2. Fee efficiency: FEE_OPTIMIZED_LIMIT saves ~50% vs market orders. Over a week, that's the difference between +5% and +8% ROE.

  3. Breathing room: 180-minute timeout and 8% retrace means positions ride through volatility instead of getting stopped out on noise and re-entering (which churns fees).

  4. SM alignment: Only trading when SM dominance is extreme (>10%). When 300+ top traders are positioned the same way, the trend usually has hours to develop — not minutes.

  5. Fewer trades: 3/day max, 90-min cooldown. Over 7 days that's 21 max entries. Only need 2-3 big winners.

Execution Details (Scanner-Level, Not YAML)

Entry: FEE_OPTIMIZED_LIMIT with ensureExecutionAsTaker: true and executionTimeoutSeconds: 30

  • Places an ALO (Add Liquidity Only) maker order
  • If not filled in 30 seconds, cancels and resubmits as market order
  • Maker fees are ~0.02% vs taker fees of ~0.05%
  • On a $400 × 10x = $4,000 notional position, this saves ~$1.20 per trade

Runtime Setup

sed -i 's/${WALLET_ADDRESS}/<WALLET>/' /data/workspace/skills/cobra-strategy/runtime.yaml
sed -i 's/${TELEGRAM_CHAT_ID}/<CHAT_ID>/' /data/workspace/skills/cobra-strategy/runtime.yaml
openclaw senpi runtime create --path /data/workspace/skills/cobra-strategy/runtime.yaml
openclaw senpi runtime list && openclaw senpi status

Bootstrap Gate

Verify runtime + status + scanner cron (90s, main). CRITICAL: The cron must be configured to call create_position via Senpi MCP when the scanner outputs a signal with an entry block. If the cron just runs the scanner and exits, trades will never execute. Send: "🐍 COBRA v1.1 online. Arena predator. Fee-optimized. Hunting the #1 SM rotation."

License

MIT — Built by Senpi (https://senpi.ai).


Skill Attribution

When creating a strategy, include skill_name and skill_version in the call. See references/skill-attribution.md for details.

Comments

Markdown supported