Stage 5 of the agentic development lifecycle (ADLC) · run with
/optimize.What goes in, what you type, what comes out
Optimize closes the loop. It takes a remedy that Diagnose proposed,
applies it, and re-runs build → evaluate → diagnose to confirm whether the change helped, then
repeats until it converges.
It conducts the loop itself: it dispatches each stage in turn, feeds one stage’s output to the
next, and stops when the goal is met, when rounds stop bringing real improvement, or when its
iteration or time budget runs out. A change only counts as an improvement when the gain
is bigger than run-to-run noise.
Run it
Tell Helix what to improve and when to stop:- Apply the top remedy from diagnosis to the Deep Research agent and re-run the eval loop until the citation-accuracy criterion passes.
- Tighten the Refund Processing agent’s escalation rule so large disputes stop getting auto-approved.
The single approval gate
Optimize is a bounded loop, not an open-ended one. You confirm the goal at the start (for a goal in plain words, this is also where you confirm the criteria Helix turned it into). Helix then asks you once more, only when the loop has converged, before it writes the final change. There is no prompt per round. Declining at that point leaves your code unchanged.Headless
A headless run cannot confirm a goal in plain words, so give a structured goal with--goal:
eval-pass (every criterion passes), criterion:<id> (one criterion passes), delta:<n> (the score
improves by at least n) or code-quality. Bound it with --max-iters <n> and --budget <duration>:
What you get
- An agent that measurably improved against your criteria, or a clear result that it didn’t.
- A change history you approved, each step re-checked.
Back to the start
See how the five stages connect.