virtuslab Using Orca
Orca (orca CLI) runs a scripted plan → code → review flow autonomously,
using its own coding/review agents. Delegate to it instead of implementing
the task yourself when that fits better.
When to use
- The task is well-defined and self-contained, with clear acceptance criteria (a feature or bugfix) — suited to an autonomous plan → code → review flow.
- NOT for exploratory/interactive work, or edits small enough to just make yourself.
How
Requires orca installed (one-line curl install — see the README's
"Getting set up" / "Orca Shell" sections).
orca run implement.sc "<task description>"
implement.sc is the default flow; run orca list to see other flows
across the project/global/built-in tiers.
Key flags:
--verbose— print a stack trace if the flow aborts.--skip-branch— continue on the current branch instead of creating a new one. Use this when the current branch already has plan/context files the flow should pick up (e.g. it was planned with a harness first). Uncommitted or untracked files are fine — no need to commit them first, the flow leaves them in place and picks them up.--honor-pin— run the flow's own pinned orca version instead of forcing this shell's.
Don't pre-create a branch: the flow creates its own, unless --skip-branch
is passed.
After it runs
Exit codes: 0 success, 1 action failure, 2 usage error — orca run
propagates the flow's own exit code. On success the flow has committed its
work on a branch; report that branch (and any PR) to the user.
If a run is interrupted, re-run the same orca run command: flows are
resumable and pick up from the last committed stage.
orca continue (list sessions with --list, resume one by selector)
reattaches to a recorded harness session, but requires a real terminal and
errors without one — don't invoke it headlessly; tell the user to run it
themselves instead.