codex▌

Spawn autonomous subagents to offload context-heavy work and preserve parent token budget. \n \n Subagents burn their own tokens and return only final results, ideal for deep research (3+ searches), codebase exploration (8+ files), multi-step workflows, and long-running operations \n Choose between mini model (gpt-5.1-codex-mini) for pure search tasks or inherit parent model for multi-step analysis, refactoring, and generation work \n Supports up to 5 parallel subagents via background shell exec

AI-powered code analysis, refactoring, and automated editing via Codex CLI with GPT-5.2. \n \n Runs codex exec and codex resume commands with configurable reasoning effort ( xhigh , high , medium , low ) and sandbox modes (read-only, workspace-write, danger-full-access) \n Defaults to GPT-5.2 model (76.3% SWE-bench performance); supports gpt-5.2-max, gpt-5.2-mini, and gpt-5.1-thinking for different complexity and cost trade-offs \n Session continuity: resume prior Codex work at any time using co

Use OpenAI Codex CLI as a read-only oracle — planning, review, and analysis only. Codex provides its perspective; you synthesize and present results to the user.

Codex is powered by OpenAI models with their own knowledge cutoffs and limitations. Treat Codex as a colleague, not an authority.

Cross-model validation using the codex binary directly. Claude writes code, Codex reviews it — different architecture, different training distribution, no self-approval bias.

GPT-5.2 Advantages: 76.3% SWE-bench (vs 72.8% GPT-5), 30% faster on average tasks, better tool handling, reduced hallucinations, improved code quality. Knowledge cutoff: September 30, 2024.

$1f

The lead orchestrates, Codex agents execute. Each agent gets one focused task. The team lead prevents file conflicts before spawning — the orchestrator IS the lock manager.

Automated code review with intelligent difficulty assessment and CHANGELOG synchronization. \n \n Automatically detects uncommitted changes or reviews the latest commit; stages untracked files and updates CHANGELOG before invoking the review \n Evaluates task complexity (file count, line changes, refactoring scope) and selects appropriate model and reasoning effort level \n Separates preparation (CHANGELOG, staging) from review execution in isolated context to reduce token waste \n Supports Go,

The script path is: