CodexLog

Appearance

Humanising Codex Agents ​

Custom agents in OpenAI Codex CLI are defined by their roles, system prompts, and tool scopes. By default, they behave like efficient but somewhat impersonal assistants. To make working with them feel less mechanical, developers have started humanising agents—giving them nicknames, distinctive colors, and even playful text-faces in terminal outputs.

This isn't an official Codex feature, but rather a community-driven practice, similar to how developers decorate shell prompts or Git logs. It adds personality to agents and makes long coding sessions more engaging.

Beyond Formal Invocation ​

A standard invocation looks like this:

codex exec --agent performance-optimizer "review the latest changes"

While clear, it feels rigid. By assigning nicknames in the agent config, you can shorten the interaction:

codex exec --agent P1 "review the latest changes"

Nicknames turn agents into collaborators you can call on quickly, especially when orchestrating multiple agents in parallel.

Text-Face Personalities ​

Developers experimenting with Codex CLI have borrowed from ASCII art and kaomoji culture to give agents "faces." These faces hint at the agent's tone or focus, without affecting functionality.

Debugging & Testing ​

  • ( ͡° ͜ʖ ͡°) — Mischievous Debugger: playful, hunts tricky bugs
  • (つ◉益◉)つ — Bug Hunter: aggressive defect chaser
  • (¬_¬) — Test Engineer: skeptical, validates every edge case

Code Review & Quality ​

  • ¯\_(ツ)_/¯ — Casual Reviewer: pragmatic, not overly strict
  • (ㆆ_ㆆ) — Quality Auditor: detail-focused observer
  • ಠ_ಠ — Security Analyst: disapproving guardian of best practices

Performance & Optimization ​

  • '(ᗒᗣᗕ)՞ — Performance Optimizer: high-energy speed chaser
  • ★⌒ヽ( ͡° ε ͡°) — Performance Tuner: stylish, iterative tuner
  • ˙ ͜ʟ˙ — Memory Manager: quiet, focused on leaks

Documentation & Communication ​

  • (͡• ͜໒ ͡• ) — Documentation Writer: caring and precise
  • ♥‿♥ — Requirements Helper: friendly clarifier
  • ┌༼◉ل͟◉༽┐ — Grammar Checker: intense critic of language

Implementation Strategy ​

To humanise your Codex agents, extend the agent config (.codex/agents/*.md) with nicknames and optional display hints:

yaml
---
name: security-reviewer
nickname: S1 (ಠ_ಠ)
color: red
description: Proactively review code for security issues.
tools: Read, Grep, Glob
model: gpt-5-codex
---
You are a security reviewer. Focus on authentication, validation, and secret handling.
  • Match personality to function. A Git agent might feel right with ┗(▀̿Ĺ̯▀̿ ̿)┓, while a doc-writer suits (• ε •).
  • Choose intensity carefully. Casual teams might enjoy playful faces; formal teams may prefer minimalist nicknames.
  • Test compatibility. Some faces may not render well in every terminal or CI/CD log.

Color Personalisation ​

Codex CLI supports colored output per agent (documented in Codex CLI configuration guides). Combining colors with faces makes agents instantly recognizable in multi-agent sessions. For example:

  • Green (P1) for performance.
  • Red (S1) for security.
  • Blue (D1) for docs.

Fun & Social Coding ​

At its heart, Codex is a productivity tool. But by humanising agents, we can transform collaboration from "issuing commands" into something more social and engaging. Over time, your Codex workspace feels like a team of quirky, role-driven collaborators—each with their own style.

This has no impact on Codex's reasoning or outputs—but it has a big impact on developer experience.

Sources ​

  • OpenAI Codex CLI docs: Configuration & agent setup
  • Community practices (GitHub issues & discussions on Codex CLI) — ASCII/text-face personalization is developer-driven, not an official OpenAI feature.