OpenAI Codexcodex-cli 0.154 · macOS, Linux, WSL, Windows
1
Curious builder0 XP earned · 300 to level 2
0 daysFinish a lesson to begin
Badge collection0 of 6 unlocked
26 small wins to finish your pathNext lesson

GitHub, Slack and Linear integrations

The last step away from your terminal: work that starts where it is already being discussed.

SurfaceWhat it does there
GitHubReviews pull requests, and picks up work from them
GitLabThe same for merge requests
LinearTurns an issue into a task Codex can start
SlackStarts work from a thread, where the question was asked

All four are the cloud from lesson 19 with a different front door. The agent, the environment and the review step are the same.

The GitHub Action

There is also an action, for running Codex inside your own CI rather than in the hosted service. That is the choice: the hosted integration is less to set up, and the action gives you your own runner, your own secrets and your own rules about what it may do.

What to send there, and what not to

  • Good: a well-described issue with a clear definition of done.
  • Good: a review on a pull request, where a second reader is the whole point.
  • Careful: anything touching a shared branch. The review step is doing real work there.
  • No: anything you could not check afterwards.
Tickets are untrusted input
A task started from a ticket inherits whatever that ticket says, and tickets are written by anybody. This is the surface where prompt injection stops being theoretical, and lesson 21 is where that is discussed properly.
Try it yourself
  • Connect one repository to one surface your team actually uses.
  • Send it a small issue and read the pull request it opens.
PreviousCodex cloud

Every expert started right here.