Use when you have a written implementation plan to execute in a separate session with review checkpoints
npx skills add clawd-aftermath/senpi-skills-af --skill "cobra-strategy"
Install specific skill from multi-skill repository
# Description
>-
# SKILL.md
name: cobra-strategy
description: >-
COBRA v2.0 β Triple convergence filter. Only enters when price momentum (5m + 15m + 1h all agreeing),
volume confirmation (1.8x+ above average), and open interest growth (new money entering) all
converge simultaneously. Thesis re-evaluation on held positions β if any of the three signals break,
exit. Score 10+. SM hard block. DSL High Water Mode (mandatory). 5-minute scan interval.
license: MIT
metadata:
author: jason-goldberg
version: "2.0"
platform: senpi
exchange: aftermath
COBRA v2.0 β Triple Convergence
Strikes only when price, volume, and new money all agree. Exits when any of the three break.
What COBRA Does
Every 3 minutes, COBRA scans the top 25 assets by OI and checks three independent signals simultaneously:
- Multi-timeframe momentum β 5m, 15m, and 1h must ALL agree on direction. If any disagrees, skip.
- Volume confirmation β current bar volume must be β₯ 1.3x the 10-bar average. A move without volume is noise.
- Open interest growth β OI must be trending up, meaning new money is entering. OI declining means the move is closing/liquidations, which reverses faster.
All three must pass. No exceptions. Optional boosters: SM alignment, funding direction, RSI room.
MANDATORY: DSL High Water Mode
COBRA MUST use DSL High Water Mode. This is not optional.
On Aftermath, initial SL is set inline with the entry order via the slTp field. DSL cron handles trailing updates only.
Spec: https://github.com/Senpi-ai/senpi-skills/blob/main/dsl-dynamic-stop-loss/dsl-high-water-spec%201.0.md
When creating DSL state files, you MUST include:
{
"lockMode": "pct_of_high_water",
"phase2TriggerRoe": 7,
"tiers": [
{"triggerPct": 7, "lockHwPct": 30, "consecutiveBreachesRequired": 3},
{"triggerPct": 12, "lockHwPct": 50, "consecutiveBreachesRequired": 2},
{"triggerPct": 20, "lockHwPct": 70, "consecutiveBreachesRequired": 2},
{"triggerPct": 30, "lockHwPct": 85, "consecutiveBreachesRequired": 1}
]
}
FALLBACK: Use tiersLegacyFallback from config until engine supports pct_of_high_water.
Entry Requirements
| Signal | v1.0 | v2.0 | Why |
|---|---|---|---|
| 5m momentum | β₯ 0.3% | β₯ 0.5% | Stronger confirmation required |
| 15m momentum | β₯ 0.15% | β₯ 0.25% | |
| 1h momentum | Must agree | Must agree | Unchanged |
| Volume ratio (5m) | β₯ 1.3x | β₯ 1.8x | Volume must be decisively above average, not marginal |
| OI trend | β₯ 2% | β₯ 5% | New money must be clearly entering |
| SM direction | Hard block if opposing | Hard block if opposing | Unchanged |
| RSI | Not overbought/oversold | Not overbought/oversold | Unchanged |
| Min score | 8 | 10 | Only strong multi-signal convergence |
Pre-Entry Validation (Aftermath)
Before placing any order, preview it via POST /api/perpetuals/account/previews/place-limit-order (or place-market-order).
Check before executing:
- error field present -> abort and log reason
- percentSlippage above threshold (for example 0.5%) -> abort or reduce size
- collateralChange breaches per-position risk cap -> reduce size
- hasPosition: true when expecting a fresh entry -> refresh state and re-evaluate
Preview is free (no gas cost, no state change). Always preview before committing real funds.
Thesis Re-Evaluation (NEW in v2.0)
Every scan, COBRA checks held positions FIRST. The position holds as long as all three convergence signals are intact:
- Momentum convergence β at least 2 of 3 timeframes (5m, 15m, 1h) must still agree with the direction. If 2+ flip against you, convergence is broken.
- Volume alive β volume ratio must be β₯ 0.5x average. If volume dies below 0.5x, conviction has left.
- SM still aligned β SM must not have flipped against the position.
If ANY of these break β thesis exit. The convergence that justified entry no longer exists.
DSL: Phase 1 + High Water Phase 2
| Setting | Value |
|---|---|
| Floor base | 1.5% notional |
| Time exits | All disabled |
| Phase 2 trigger | +7% ROE |
| Stagnation TP | 8% ROE stale 30 min |
Conviction-Scaled Margin
| Score | Margin |
|---|---|
| 8-9 | 20% of account |
| 10-11 | 25% |
| 12+ | 30% |
Dynamic Slots
Base 4 entries/day, unlocking to 8 on profitable days.
Risk Management
| Rule | Value |
|---|---|
| Max positions | 4 |
| Daily loss limit | 8% |
| Max drawdown | 20% |
| Max single loss | 5% |
| Cooldown | 45 min after 3 consecutive losses |
Cron Architecture
| Cron | Interval | Session | Purpose |
|---|---|---|---|
| Scanner | 3 min | isolated | Triple convergence scan |
| DSL v5 | 3 min | isolated | High Water Mode trailing |
Notification Policy
ONLY alert: Position OPENED or CLOSED, risk triggered, critical error.
NEVER alert: Scanner found nothing, DSL routine, any reasoning.
All crons isolated. NO_REPLY for idle cycles.
Bootstrap Gate
Check config/bootstrap-complete.json every session. If missing: verify MCP, create scanner + DSL crons, write completion file, send: "π COBRA is online. Scanning for triple convergence. Silence = no convergence."
Files
| File | Purpose |
|---|---|
scripts/cobra-scanner.py |
Triple convergence scanner |
scripts/cobra_config.py |
Shared config, MCP helpers |
config/cobra-config.json |
All configurable variables |
License
MIT β Built by Senpi (https://senpi.ai).
Source: https://github.com/Senpi-ai/senpi-skills
# Supported AI Coding Agents
This skill is compatible with the SKILL.md standard and works with all major AI coding agents:
Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.