You want an agent to run as part of CI. What do you need to decide?
Why C is the answer
What invokes it, and what scope and permissions it has in that run
Running an agent in CI means deciding what invokes it, and what scope and permissions it has in that run.
In detail
- Decide what invokes the agent, and what scope and permissions it has.
- Write down the trigger and the scope in the workflow.
Use case
An illustrative scenario, not a real incident.
A team adds an agent to CI. It decides that a pull request triggers it, with read access to code and write access to comments only.
See it in different ways
stateDiagram-v2 [*] --> Trigger Trigger --> Setup: copilot-setup-steps Setup --> Runner: standard, larger or self-hosted Runner --> Working Working --> DraftPullRequest DraftPullRequest --> [*]
mindmap
root(("Agent in CI"))
n1["Trigger"]
n2["What invokes it"]
n3["Scope"]
n4["What it works on"]
n5["Permissions"]
n6["What it may change"]Key point: Decide the trigger, scope and permissions.
Why the other options are wrong
- A. How many comments to add to the workflow file for future readers
- Comments do not define what the agent can do.
- B. Whether the workflow should be renamed to mention artificial intelligence
- The workflow name has no effect.
- D. Which of the team's members should watch the job until it ends
- CI should run without someone watching.