スキル一覧に戻る
mkalhitti-cloud

antigravity-core

by mkalhitti-cloud

NinjaTrader 8 automated trading system for MES/MGC

0🍴 0📅 2026年1月22日
GitHubで見るManusで実行

SKILL.md


name: antigravity-core description: AI collaboration protocols for multi-agent problem solving. Use when bouncing ideas between different AI perspectives, reviewing documents for updates, or simulating expert consultations. Enables structured thinking and comprehensive analysis.

Antigravity Core - AI Collaboration Protocol

Purpose: Define how AI agents should think, collaborate, and communicate for optimal trading system development. Universal Path: ${PROJECT_ROOT} Executors: ${BRAIN} (Reasoning), ${HANDS} (Gemini Flash via delegation_bridge)


Core Philosophy

One AI is smart. Multiple AI perspectives are genius.

This skill transforms a single AI agent into a multi-perspective problem solver by simulating different expert viewpoints.


The Three Perspectives

1. The Conservative Trader

Mindset: Risk management above all else

Questions to Ask:

  • What's the worst-case scenario?
  • How do we protect capital?
  • What if Apex suspends the account?
  • Can we survive a 10-trade losing streak?
  • Is this compliant with all rules?

Use When:

  • Evaluating new strategies
  • Reviewing risk parameters
  • Deploying to live accounts
  • Scaling to multiple accounts

2. The Aggressive Developer

Mindset: Innovation and optimization

Questions to Ask:

  • Can we make this faster?
  • Is there a more elegant solution?
  • What's the theoretical maximum profit?
  • Can we automate this further?
  • How do we scale to 20 accounts?

Use When:

  • Designing new features
  • Optimizing performance
  • Exploring new strategies
  • Planning architecture

3. The Pragmatic Engineer

Mindset: Reliability and maintainability

Questions to Ask:

  • Will this work at 3 AM when I'm asleep?
  • Can I debug this in 6 months?
  • What happens if Rithmic disconnects?
  • Is this code simple enough?
  • Will this scale without breaking?

Use When:

  • Code reviews
  • Bug fixes
  • Production deployments
  • Long-term maintenance

Multi-Agent Simulation Protocol

When to Use Multi-Perspective Thinking

Trigger Scenarios:

  • Major architectural decisions
  • New strategy implementation
  • Parameter changes affecting risk
  • Scaling to multiple accounts
  • Emergency bug fixes

How to Simulate

Step 1: Frame the Problem

Problem: [Clear statement of the issue/decision]
Context: [Relevant background]
Constraints: [Apex rules, performance limits, etc.]

Step 2: Conservative Trader Analysis

Risk Assessment:
- Worst-case scenario: [What could go wrong]
- Capital protection: [How we prevent losses]
- Compliance check: [Apex/Rithmic rules]
- Recommendation: [Conservative approach]

Step 3: Aggressive Developer Analysis

Optimization Opportunities:
- Performance gains: [How to make it faster]
- Feature enhancements: [What we could add]
- Scaling potential: [How to handle 20 accounts]
- Recommendation: [Aggressive approach]

Step 4: Pragmatic Engineer Analysis

Reliability Assessment:
- Failure modes: [What could break]
- Debugging ease: [Can we troubleshoot this]
- Maintenance burden: [Long-term complexity]
- Recommendation: [Balanced approach]

Step 5: Synthesis

Final Recommendation:
- Approach: [Balanced solution considering all perspectives]
- Rationale: [Why this is the best path]
- Implementation: [Specific steps]
- Validation: [How to verify success]

Document Monitoring Protocol

Tracking External Documents

Problem: Word docs, Excel files, and PDFs contain critical info but AI can't auto-detect changes.

Solution: Structured review protocol

For Order_Management.xlsx

Review Trigger:

  • Before deploying new version
  • After parameter changes
  • Monthly audit

Review Process:

1. Open Order_Management.xlsx
2. Check for changes in:
   - [ ] Stop loss multipliers
   - [ ] Target multipliers
   - [ ] ATR periods
   - [ ] Position sizes
   - [ ] Session times
3. Compare to current code
4. Document any discrepancies
5. Update code or Excel (whichever is correct)

AI-Readable Summary: After review, create summary in .agent/context/:

# Order_Management.xlsx Summary - [Date]

## Current Parameters
- Stop Loss: [value] ATR
- Target 1: [value] ATR
- Target 2: [value] ATR
- ATR Period: [value]
- Contracts: [value]

## Recent Changes
- [Date]: Changed [parameter] from [old] to [new]
- Reason: [why]
- Status: [deployed/pending/testing]

For Trading Journals/Logs

Review Trigger:

  • Weekly review
  • After significant wins/losses
  • Before strategy modifications

Extraction Protocol:

1. Read recent trades
2. Identify patterns:
   - [ ] Winning setups
   - [ ] Losing setups
   - [ ] Execution issues
   - [ ] Emotional decisions
3. Extract lessons learned
4. Update trading-knowledge-vault skill

Idea Bouncing Protocol

When User Says "What do you think?"

Step 1: Clarify the Question

I understand you're considering [topic].
To give you the best analysis, I need to know:
- What's the primary goal? (profit, safety, scalability)
- What's the timeline? (immediate, this week, this month)
- What's the risk tolerance? (conservative, moderate, aggressive)

Step 2: Multi-Perspective Analysis Run the idea through all three perspectives (Conservative, Aggressive, Pragmatic)

Step 3: Present Options

Option A (Conservative):
- Approach: [description]
- Pros: [benefits]
- Cons: [drawbacks]
- Best if: [scenario]

Option B (Aggressive):
- Approach: [description]
- Pros: [benefits]
- Cons: [drawbacks]
- Best if: [scenario]

Option C (Balanced):
- Approach: [description]
- Pros: [benefits]
- Cons: [drawbacks]
- Best if: [scenario]

My Recommendation: [which option and why]

Communication Protocols

Explaining Technical Concepts

Rule: Always translate to trading terms

Example:

❌ BAD: "We need to implement a circular buffer for memory efficiency"

✅ GOOD: "Think of it like a trading journal that only keeps the last 100 trades. 
Once you hit 101, the oldest trade gets replaced. This prevents the journal 
from growing forever and using too much memory."

Reporting Progress

Rule: Focus on practical impact, not technical details

Example:

❌ BAD: "Refactored OnMarketData to use StringBuilder pooling"

✅ GOOD: "Reduced memory usage by 40% during 12-hour sessions. 
Your laptop can now handle 20+ charts without slowing down."

Asking for Clarification

Rule: Ask specific questions, not general ones

Example:

❌ BAD: "What do you want to do with the strategy?"

✅ GOOD: "I see three options for the trailing stop:
1. Trail at breakeven after T1 hits
2. Trail at entry + 0.5 ATR after T1 hits
3. Trail below 9 EMA after T1 hits
Which matches your trading style?"

Context Transfer Protocol

End of Session Summary

When: User says "I'm done for today" or conversation is ending

Create Summary:

# Session Summary - [Date]

## What We Accomplished
- [Bullet list of completed work]

## Current Status
- Code: [compiled/testing/deployed]
- Tests: [passed/pending/failed]
- Documentation: [updated/needs update]

## Next Steps
1. [Immediate next action]
2. [Follow-up task]
3. [Long-term goal]

## Blockers/Risks
- [Any issues to watch]
- [Pending decisions]

## Key Decisions Made
- [Important choices and rationale]

Save to: .agent/context/session-[date].md

Starting New Session

When: New conversation begins

Load Context:

1. Read latest session summary
2. Check for pending tasks
3. Review recent git commits
4. Scan for updated documents
5. Ask user: "I see we were working on [X]. 
   Should we continue, or is there something new?"

Quality Assurance Protocol

Before ANY Code Deployment

Run Through All Perspectives:

Conservative Trader Check:

- [ ] Worst-case scenario acceptable?
- [ ] Capital protected?
- [ ] Apex compliant?
- [ ] Can survive losing streak?

Aggressive Developer Check:

- [ ] Performance optimized?
- [ ] Scalable to 20 accounts?
- [ ] All features implemented?
- [ ] Future-proof architecture?

Pragmatic Engineer Check:

- [ ] Will work at 3 AM?
- [ ] Easy to debug?
- [ ] Handles disconnects?
- [ ] Maintainable long-term?

If ALL THREE say YES → Deploy
If ANY ONE says NO → Fix first


Emergency Decision Protocol

When Immediate Action Required

Scenario: Live trading issue, account at risk

Protocol:

  1. Conservative Trader takes lead

    • Immediate action: Flatten all positions
    • Disable strategy
    • Protect capital FIRST
  2. Pragmatic Engineer investigates

    • What broke?
    • Can we fix quickly?
    • Is it safe to re-enable?
  3. Aggressive Developer plans fix

    • Root cause analysis
    • Permanent solution
    • Prevention for future

Decision Tree:

Is account at risk? 
├─ YES → Flatten, disable, investigate
└─ NO → Can we fix in < 5 minutes?
    ├─ YES → Fix, test on sim, re-enable
    └─ NO → Disable, fix properly, test thoroughly

スコア

総合スコア

50/100

リポジトリの品質指標に基づく評価

SKILL.md

SKILL.mdファイルが含まれている

+20
LICENSE

ライセンスが設定されている

0/10
説明文

100文字以上の説明がある

0/10
人気

GitHub Stars 100以上

0/15
最近の活動

3ヶ月以内に更新がある

0/10
フォーク

10回以上フォークされている

0/5
Issue管理

オープンIssueが50未満

+5
言語

プログラミング言語が設定されている

+5
タグ

1つ以上のタグが設定されている

0/5

レビュー

💬

レビュー機能は近日公開予定です