Codex and ChatGPT¶

Codex CLI¶

Command syntax verified with Codex CLI 0.167.0:

codex plugin marketplace add uncscode/particula
codex plugin add particula-helper@particula
codex plugin list --marketplace particula --json

Start a new session in your project, then ask it to use particula-setup and your existing .venv. The repository marketplace advertises the plugin without automatically enabling it for every contributor.

For a release-pinned installation, replace RELEASE_TAG with an actual tag:

codex plugin marketplace add uncscode/particula --ref RELEASE_TAG
codex plugin add particula-helper@particula

Refresh and reinstall to pick up an updated plugin version:

codex plugin marketplace upgrade particula
codex plugin add particula-helper@particula

Remove only this plugin, optionally followed by its marketplace:

codex plugin remove particula-helper@particula
codex plugin marketplace remove particula

For local development, register your checkout instead of the GitHub source:

codex plugin marketplace add /absolute/path/to/particula
codex plugin add particula-helper@particula

Do not register a local checkout and the remote repository under the same marketplace name simultaneously. Remove the old marketplace before switching.

ChatGPT and desktop packages¶

In a desktop host that supports repo marketplaces, open its plugin browser, select the particula marketplace, and install Particula Helper. Restart or start a new chat after installing. Where custom plugin uploads are supported, use particula-helper-codex.zip from a Particula release.

The ZIP uses the portable Agent Plugins manifest and includes a Codex compatibility manifest. Repository distribution is separate from publication in the public plugin directory; no public directory listing is implied. Web/account upload availability depends on the host and account. This release does not claim that the desktop upload path has been tested.

Skills can provide instructions anywhere they load, but simulations require Python execution and filesystem access. A web-hosted interpreter is not your local project environment. Never assume it can use your local .venv.

See OpenAI plugin packaging and marketplace documentation and the shared first-run guide.