Skip to content

teaceo-debug/DEEP6 / .claude/skills/deep6-edge-research

deep6-edge-research

Unverified

DEEP6 Edge Research Skill

Purpose

This skill equips agents to conduct rigorous, reproducible backtesting research on the DEEP6 signal engine — specifically DOM/order flow signals on NQ futures. It covers the full research loop: data access → signal collection → attribution analysis → edge identification → hypothesis testing.

When to load this skill

Load this skill when any agent is asked to:

  • Find which DEEP6 signals have real edge
  • Test a signal or combination hypothesis
  • Run backtests on DOM / footprint / order flow signals
  • Build new signal detectors and measure their alpha
  • Run signal attribution after downloading Databento MBO data
  • Answer "does X signal work?" or "what's the best time of day for Y signal?"
  • Compare signal performance across market regimes

Skill entry point

  1. Read knowledge.md — master framework, data paths, signal taxonomy
  2. Read the relevant playbook for your task:
    • playbooks/signal-attribution.md — run and interpret attribution
    • playbooks/mbo-backtesting.md — MBO-specific testing (Databento data)
    • playbooks/dom-signal-dev.md — building and testing new DOM signals
    • playbooks/hypothesis-testing.md — testing specific edge hypotheses
  3. Read data/signal-taxonomy.md — full 44-signal reference with categories

Multi-agent parallelization

This skill is designed for PARALLEL agent workstreams. Each agent takes one category:

AgentFocusCommand
Agent AAbsorption + Exhaustionpython scripts/signal_analyze.py --category absorption
Agent BImbalance + Deltapython scripts/signal_analyze.py --category imbalance
Agent CVolume Profile + Auctionpython scripts/signal_analyze.py --category volume_profile
Agent DDOM engines (ENG_02-05)python scripts/signal_analyze.py --signal ENG_02,ENG_03,ENG_04,ENG_05
Agent ETrap + Vol Patternspython scripts/signal_analyze.py --category trapped
Agent FPair synergy researchpython scripts/signal_analyze.py --no-pairs then pair drill-down
Agent GTime-of-day researchQuery signal_events.csv directly, group by hour
Agent HRegime-conditioned edgeFilter signal_events.csv by trend/mean-reversion periods

Knowledge Base (load for deep research tasks)

FileContentsWhen to load
knowledge/mbo-strategy-taxonomy.mdAll 22 manipulation patterns with MBO signatures, detection algorithms, counter-strategiesAny spoof/iceberg/layering/absorption work
knowledge/mbo-algorithms.mdComplete Python implementations: OFI, Kyle's λ, Hawkes, VPIN, queue tracker, iceberg detector, spoof detector, feature vectorBuilding or debugging detectors
knowledge/institutional-architecture.mdFull 9-layer system blueprint: data → book → signals → levels → features → LLM → classifiers → decisions → executionArchitecture decisions, new component design
knowledge/nq-strategy-library.mdNQ-specific strategies: ORB, VWAP reversion, stop sweeps, options-futures interaction, HFT signaturesStrategy development, hypothesis generation

Base path

C:\Users\Tea\DEEP6\.claude\skills\deep6-edge-research\

add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution.Stars: 114kUpdated Claude CodeCodexVerifiedInstall
add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation.Stars: 114kUpdated Claude CodeCodexVerifiedInstall
diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope.Stars: 114kUpdated Claude CodeCodexVerifiedInstall
fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks.Stars: 114kUpdated Claude CodeCodexVerifiedInstall