- Home
- Skills
- Agents & Orchestration
- DualAgent Nexus Pro: Multi-Agent Cascade & Brain Selector
More screenshots
Works with the AI Agents you already use
DualAgent Nexus Pro: Multi-Agent Cascade & Brain Selector
Interactive agents like Claude Code, Cursor, Codex, and Gemini CLI excel at high-level reasoning, system design, and iterative architecture.
$19
DualAgent Nexus Pro: Multi-Agent Cascade & Brain Selector
Example session with this skill installed
Evaluate this task: I need to generate 20 unit tests for the billing service and update the schema in 6 related files. Should I delegate this?
- Read your context and instructions
- Compiled the dualagent nexus pro
I have analyzed the task requirements and current resource availability.
Task Triage and Delegation Plan
| Metric | Value |
|---|---|
| Task Type | Multi-file Migration / Test Gen |
| Estimated Volume | ~450 lines of code |
| Routing Decision | DELEGATE |
| Recommended Pool | agy-claude (Tier: Heavy) |
| Est. Token Offload | ~12,500 tokens |
The task meets the threshold for autonomous delegation. I will proceed by creating a Git checkpoint and dispatching the worker.
# 1. Create safety boundary
git add -A && git commit -m "checkpoint: pre-billing-migration"
# 2. Dispatch the cascade
python scripts/cascade.py --role executor --tier heavy --dir src/services/billing --yolo "Generate 20 unit tests and update schema across 6 files"
Next steps
- Review the
===DIGEST===block once the worker completes. - Run
npm test src/services/billingto verify the generated tests. - Inspect the changes using
git diffbefore final commit.
Connects securely to your tools. The creator never sees your data.
About this skill
Every software engineer using modern AI coding agents inevitably encounters the same painful barrier: the primary interactive session runs out of tokens or degrades under severe context window bloat. Interactive agents like Claude Code, Cursor, Codex, and Gemini CLI excel at high-level reasoning, system design, and iterative architecture. However, when an agent attempts to author twenty unit tests, execute an eight-file framework migration, or generate repetitive CRUD boilerplate directly inside its conversational window, thousands of tokens of raw code, compiler logs, and intermediate diffs flood the context. This induces the "Lost in the Middle" degradation effect, causes instruction drift, and prematurely burns through expensive monthly subscription quotas.
DualAgent Nexus Pro v1.2.0 solves this architectural challenge through out-of-band execution, hybrid multi-pool cascades, and explicit role-based agent orchestration. By decoupling the interactive conductor (the Brain) from mechanical implementation, Nexus Pro offloads token-heavy coding tasks to prioritized pools of headless worker CLIs and direct pay-per-token API endpoints (Anthropic, DeepSeek, OpenAI, OpenRouter, and local Ollama servers). The conductor remains unpolluted, receiving only a machine-verifiable, 300-token ===DIGEST=== summarizing status, modified files, commits, and verification test results.
When a worker pool encounters rate limits or capacity depletion (HTTP 429 or Exit Code 10), Nexus Pro automatically writes a local cooldown lock, emits structured telemetry, and cascades execution to the next configured quota pool without human intervention. Standardized role presets (claude-brain, claude-brain-codex, codex-brain, gemini-brain, api-budget, local-first) configure optimal waterfall priorities with a single command, while centralized MCP management (mcp_manager.py) non-destructively synchronizes tools across seven host platforms. A unified CLI usage dashboard tracks live 5-hour subscription limits, reset countdowns, and API dollar budgets.
Nexus Pro is completely local-first and dependency-free: it runs on the Python 3.9+ standard library with no pip packages. Network access is confined to a single audited file (api_worker.py) that communicates strictly with your configured API endpoints when an API pool is selected. Delivered with 20 executable scripts, 9 slash commands for 4 agents (36 command files), 14 architectural reference guides, 29 configuration templates, and 8 complete worked examples (159 total files).
New in v1.2: choose your Brain, keep every agent in sync
Make Claude, Codex, or Gemini the conductor with one command and run Claude Code and Codex CLI as executors, each on its own account (per-pool CLAUDE_CONFIG_DIR / CODEX_HOME). Hooks record every Claude Code and Codex turn in a local per-project brain journal: executors receive that context automatically, and when one subscription runs out, nexus brain handoff --to codex lets another AI take over exactly where the conversation left off. Fully tested on Windows, macOS, and Linux (43 offline scenarios).
Highlights
- Choose Your Brain (v1.2): Make Claude, Codex, or Gemini the conductor with one command (
roles.py brain codex) or theclaude-brain-codex/codex-brainpresets. - Shared Brain Journal (v1.2): Hooks record every Claude Code and Codex turn per project; executors receive that context and
nexus brain handoff --to codexlets another AI take over when a subscription runs out. - Configurable Multi-Pool Quota Waterfall: Chain external subscriptions, secondary CLI groups, and pay-per-token API pools with automatic 30-minute cooldown locks upon HTTP 429 or Exit Code 10.
- Context Shield via Strict Digest Contract: Keep your main agent's context small by returning only a machine-validated
===DIGEST===block (~300 tokens) back to the conductor. - Decoupled Roles & Standardized Presets: Assign discrete Brain, Router, Executor, and Reviewer roles with drop-in presets (
claude-brain,gemini-brain,api-budget,local-first) viaroles.py. - Direct Pay-per-Token API Pools: Connect to DeepSeek, OpenAI, Anthropic, or local Ollama servers with automated budget tracking and safe
--write-filespath traversal guards. - Centralized Cross-Host MCP Management: Maintain a single source of truth (
mcp.servers.json), run health diagnostics (doctor), and dry-run sync tool connections across 7 major host platforms. - Unified Live Usage & Quota Monitor: Inspect rolling 5-hour and 7-day subscription limits, reset countdowns, worker availability, and API dollar balances on a single terminal screen (
usage.py). - Native Host Slash Command Bundles: Bundles 9 pre-built slash commands (
/nexus-route,/nexus-delegate,/nexus-brain, etc.) across Claude Code, Gemini CLI, Codex, and Cursor. - Deterministic Routing with Heuristic Fallback: Compare inline token costs against delegation thresholds (
route_task.py), with optional LLM classification that falls back to the heuristic offline. - Verifiable Single-File Network Policy: Network code is strictly confined to
api_worker.py(stdliburllib.request); all telemetry stays local, and keys are read exclusively from environment variables. - Complete Package: Delivered with 20 scripts, 9 slash commands for 4 agents, 14 technical manuals, 29 templates, and 8 complete worked examples checked by a bundled 43-test self-test (
selftest.py).
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Requires Git. Needs Python 3.9+. Best for tasks over 50 lines. Primary host must support shell execution to run the cascade scripts.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed today
- Accounts
- Claude CodeorCodex CLIorGemini CLIorAntigravity agyoran LLM API key
- Software
Python
Gittsc
Node.js@modelcontextprotocol/server-filesystem