Hermes Codex Imagegen Skill

What Hermes Codex Imagegen Skill does

Hermes Codex Imagegen Skill is a Hermes Agent skill for generating images through a separate Codex CLI process that is logged in with ChatGPT or OpenAI Codex OAuth. It exists for the case where Hermes's native image backend is unavailable but codex is installed and already authenticated. The skill runs codex exec in its own process, asks Codex to use its built-in imagegen skill, and captures the Codex session id.

Afterwards it re-verifies the real PNG output under ~/.codex/generated_images/ from the Hermes side, and the README warns that bwrap or sandbox inspection errors do not always mean generation failed. It is a Codex CLI workaround, not an API integration: it does not rewire Hermes's native image_generate tool, does not call the OpenAI Images API, and still needs a working Codex login. The README suggests copying a verified file into your project rather than leaving it under .codex. The repository holds a single SKILL.md with English and Korean documentation, and the skill slug is codex-imagegen-via-chatgpt-oauth.

Key features

• Runs codex exec in a separate process instead of touching Hermes's image backend
• Asks Codex to use its built-in imagegen skill
• Captures the Codex session id
• Re-verifies the generated PNG under ~/.codex/generated_images/
• Documents that bwrap or sandbox errors do not always mean the image failed
• English and Korean documentation

When to use it

• Generating an image from Hermes when no native image backend is configured
• Reusing an existing ChatGPT subscription login for image generation instead of a separate API key
• Checking that a Codex-generated image file exists before copying it into a project

Who it is for: Hermes Agent users who already have a working Codex CLI login and want image generation without setting up a separate image API.

How it fits with Hermes Agent

Built as a Hermes Agent skill (SKILL.md) that shells out to the Codex CLI and leaves Hermes's native image_generate tool unchanged.

Requirements: Hermes Agent, Codex CLI installed and on PATH, an active ChatGPT or OpenAI Codex OAuth login, and a Linux or macOS shell where codex exec can run

Note: It is a workaround that depends on a separate, logged-in Codex CLI rather than an integration with Hermes's native image backend.

FAQ