From nexus-agents
Delegates code generation tasks to Codex CLI for implementing features, generating tests, refactoring code, or bulk changes. Auto-triggers on 'delegate to codex' or 'use codex'.
npx claudepluginhub williamzujkowski/nexus-agentsThis skill is limited to using the following tools:
<!-- CANONICAL SOURCES:
Delegates complex code generation, refactoring, architectural analysis, and review tasks to OpenAI's Codex CLI (GPT-5.3-codex models) via safe workflows with sandboxing and approvals. Activates on explicit triggers like 'use codex' or 'codex exec'.
Delegates repetitive multi-file coding tasks like batch edits, boilerplate generation, large refactors, and test scaffolding to Codex CLI under Claude supervision.
Delegates coding tasks (debug, implement feature, refactor) to OpenAI Codex CLI via `codex exec`. Skips Node runtime overhead; Claude verifies output. Use for direct, fast execution.
Share bugs, ideas, or general feedback.
Full documentation:
| Metric | Value | Notes |
|---|---|---|
| Model version | gpt-5.2-codex | Latest research preview |
| Latency (simple) | 3-4 seconds | Single function generation |
| Latency (complex) | 5-6 seconds | Interface + class boilerplate |
| Token overhead | ~900-1050 per session | Fixed session initialization cost |
| Default sandbox | read-only | Safe for code generation |
| Code quality | Production-ready | Proper generics, types, error handling |
Codex excels at:
| Task Type | Examples |
|---|---|
| Code Generation | Implement functions, create classes, add features |
| Test Generation | Write unit tests, integration tests, test fixtures |
| Refactoring | Extract methods, rename variables, restructure |
| Code Completion | Complete partial implementations, fill in stubs |
| Bulk Code Changes | Apply patterns across multiple files |
Do NOT use for: Architecture decisions, security review, documentation-heavy tasks.
Use the delegate_to_model MCP tool for intelligent routing:
delegate_to_model(task: "Implement a binary search function in TypeScript")
The CompositeRouter automatically routes to Codex when:
For explicit Codex invocation, use codex exec for non-interactive mode:
# Basic code generation (non-interactive)
codex exec "Implement a binary search function in TypeScript"
# Via nexus-agents CLI
nexus-agents orchestrate "Implement sorting algorithm" --cli=codex
# Explicit model selection (default is gpt-5.2-codex)
codex exec -m gpt-5.2-codex "Implement a debounce function"
# Full autonomous mode (dangerous - writes files)
codex exec --full-auto "Add error handling to all functions in utils.ts"
Important: Use codex exec NOT codex -p. The -p flag is for interactive sessions.
Check routing decisions before execution:
nexus-agents routing-audit "Implement authentication middleware" --format=json
The router analyzes tasks using these signals:
| Signal | Boost Codex When |
|---|---|
codeGeneration | Keywords: implement, write, create |
reasoningSimple | No "design", "architect", "analyze" |
budgetSensitive | Keywords: quick, simple, straightforward |
technicalContext | References to APIs, frameworks, libs |
delegate_to_model(task: "Write Jest tests for the UserService class covering:
- createUser with valid input
- createUser with invalid email
- getUserById with existing user
- getUserById with non-existent user")
delegate_to_model(task: "Implement a rate limiter middleware for Express.js that:
- Uses sliding window algorithm
- Supports configurable limits per route
- Returns 429 status when limit exceeded")
delegate_to_model(task: "Refactor the payment processing module to:
- Extract validation logic into separate functions
- Replace callbacks with async/await
- Add TypeScript types for all parameters")
delegate_to_model(task: "Add JSDoc comments to all exported functions in src/utils/")
Task Input
|
v
TaskAnalyzer (profile task)
|
v
BudgetRouter (filter by constraints)
|
v
TopsisRouter (rank: quality 50%, cost 30%, latency 20%)
|
v
LinUCBBandit (contextual learning)
|
v
Codex selected if: high codeGeneration + low reasoningComplexity
| Capability | Score | Notes |
|---|---|---|
| Code Generation | 0.95 | Primary strength |
| Context Window | 400K | Large file support |
| Cost Efficiency | High | Lower than Claude for code tasks |
| Latency | 3-6s | Fast for focused code tasks |
tsc --noEmit on TypeScript before integrationIf Codex is unavailable:
# nexus-agents.yaml
routing:
enableTopsisRanking: true
enableLinUCBSelection: true
topsis:
qualityWeight: 0.5
costWeight: 0.3
latencyWeight: 0.2
# Check Codex availability
nexus-agents doctor
# View routing decision for task
nexus-agents routing-audit "task description" --verbose
# Show bandit learning stats
nexus-agents routing-audit "task" --bandit-stats
| Excuse | Counter |
|---|---|
| "Codex is fastest, always use it" | Codex is fastest for code generation. For research/docs, Gemini wins on context; for design, Claude wins. Match the CLI to the task type. |
| "I'll route everything through Codex" | Performance floors fire when one CLI is overweighted. Spread load per the routing system (CompositeRouter). |
| "Skip the model probe" | Required after init (#1402). Without it, you may try to use a model the CLI doesn't expose. |
| "Codex failed, retry with same params" | Transient retry already runs (#1531). If it failed twice, the failure is real — fall back to a different CLI per the routing chain. |
CompositeRouter.route() — no direct adapter instantiation