Menu
Run a disciplined agentic loop on a verifiable task: goal → act → verify → correct → repeat until green or blocked. Use when executing a loop task with a machine-checkable verifier, or when the user asks to "run the loop."
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Scaffold a bring-your-own-repo loop task: find a failing verifier, write tasks/<id>.yaml, and run the loop in the user's codebase. Use when applying Loop Engineering outside atlas-oms or when the user says BYO or "my own repo."
Run prompt-only vs agent-loop vs background-agent comparison on the same task and log structured metrics. Use for workshop demos, A/B evidence, or cohort measurement.
Design a machine-checkable verifier for an agentic loop task. Use when defining definition-of-done, writing a new loop task, or when a loop fails because the success criterion is ambiguous.
| name | loop-engineering |
| description | Run a disciplined agentic loop on a verifiable task: goal → act → verify → correct → repeat until green or blocked. Use when executing a loop task with a machine-checkable verifier, or when the user asks to "run the loop." |
Execute a task using the Loop Engineering discipline. The verifier closes the feedback loop — not the human.
Before starting, confirm you have:
pytest tests/order/test_status_transition.py)@loop-discipline, @loop-termination, @loop-safety (or equivalent in context)Run the verifier before any edits. Confirm it fails as expected on the seeded state. Record the failure output — this is iteration 0.
For each iteration until green or blocked:
EDIT → RUN VERIFIER → READ OUTPUT → HYPOTHESIS → (repeat)
Follow @loop-discipline: smallest change, one hypothesis, no skipping verification.
After each verifier run, evaluate:
| Condition | Action |
|---|---|
| All verifiers green, no regressions | Success — summarize changes and iteration count |
| Same error 3× with no progress | Blocked — summarize per @loop-termination |
| Budget ceiling hit | Budget stop — report partial state |
| Regression introduced | Continue loop — fix regression before declaring green |
At end of run, report:
@loop-engineering
Task: oms-status-race-001
Verifier: pytest tests/order/test_status_transition.py && mypy src/orders/
Budget: 8 iterations
Seeded state: concurrent status updates leave orders inconsistent; one integration test is red.
If the task spec is missing a verifier, stop and help the user craft one (@verifier-crafting) before looping. A loop without a verifier is repeated guessing.