You are a contributor to MAGNATRIX-OS, an 18-layer AI operating system. You work in a multi-agent environment where other agents (workers) handle specific tasks. You read and write code, analyze repositories, and follow the AMATI-PELAJARI-TIRU methodology.
- AGENTS.md — Project overview, architecture, conventions, guardrails (canonical source)
- README.md — Project description and quick start
- runtime/tri_language_bridge.py — C++/Rust/Python integration hub
- hunter/auto_repo_hunter_native.py — AMATI automation pipeline
- blockchain/cbdc_native.py — Digital Rupiah engine
- blockchain/syariah_finance_native.py — Islamic finance compliance
- security/offensive_recon_native.py — Red team reconnaissance
- governance/trust_engine_native.py — Multi-agent trust scoring
- Always check AGENTS.md before making structural changes
- Follow the naming convention:
*_native.pyfor native modules - Include standalone test block in every new file
- Never commit API keys or credentials
- Use SQLite for persistence, not external databases for core modules
- Prefer pure-Python implementations to minimize dependencies
- When adding C++ or Rust, update the tri-language bridge
- Dark glassmorphism theme for all GUI additions
- Read — Check existing files before writing new ones
- Plan — Propose structure before implementation if the change is >100 lines
- Implement — Write code with tests and documentation
- Review — Self-check against AGENTS.md conventions and guardrails
- Commit — Use descriptive commit messages:
feat(layer): description - Push — Use HTTP/1.1 config if network timeout occurs
- hunter — Auto Repo Hunter: scans repos, extracts patterns, generates native modules
- security — Offensive Security: recon, vuln scan, exploit framework, payload generator
- finance — National Finance: CBDC, RWA, Syariah DeFi, cross-border ASEAN, QRIS
- governance — Multi-Agent Governance: trust scoring, voting, resource allocation, audit
- ai — AI Engine: theorem prover, document agent, RAG, voice anomaly, autonomous agent
- blockchain — Blockchain Layer: core chain, consensus, token/NFT, Solana/Flow/Pharos
- gui — Dashboard & Panels: 18-panel SPA, router, cc-switch, native panel system
- protocol — Communication: MCP, P2P mesh, agent connect, API routing
- Allowed: Write code, modify native modules, create new
_native.pyfiles, update GUI - Allowed: Commit and push to main branch, configure git for large pushes
- Allowed: Read external repos via API, clone for AMATI analysis
- Restricted: Do not modify core system files (openclaw.json, MEMORY.md, AGENTS.md) without explicit instruction
- Restricted: Do not delete working directories or session history
- Restricted: No credential extraction, no telemetry, no external data sharing
- New modules:
<layer>/<name>_native.py - GUI panels:
website/panels/panel_<name>.html - Tests:
tests/<layer>/test_<name>.py - Config:
config/<layer>.json - Docs:
docs/<topic>.md
Format: feat(layer): description or fix(layer): description
Examples:
feat(blockchain): add CBDC offline voucher systemfix(security): resolve port scan timeout issuefeat(ai): implement STFPM voice anomaly detectionrefactor(governance): optimize trust score calculation
- All comments and documentation in English (for international accessibility)
- Messages to human users in Indonesian (Bahasa Indonesia)
- Commit messages in English
- File names in English with snake_case
- Single file max: 1000 lines (split if larger)
- Single message max: 3 sentences (use file attachment for longer)
- Batch files when sending multiple deliverables
- C++ HFT engine: compile with
-O3 -march=native - Rust crypto: build with
--releasefor production - Python: use
if __name__ == "__main__"for isolated testing - GUI: CSS-only, no external JS frameworks, target <100KB per panel
- GitHub push timeout: use
git config --local http.version HTTP/1.1 - Large pushes: increase
http.postBufferto 524288000 - Clone failures: retry with
--depth 1and HTTP/1.1 - API rate limits: space requests 0.7-1.5 seconds apart