Patterns for continuous autonomous agent loops with quality gates, evaluations, and recovery controls.
Scanned 6/4/2026
Install via CLI
openskills install lidge-jun/cli-jaw-skills---
name: continuous-agent-loop
description: Patterns for continuous autonomous agent loops with quality gates, evaluations, and recovery controls.
---
# Continuous Agent Loop
Patterns for structuring long-running agent workflows with loop selection, quality gates, and failure recovery.
## Loop Selection
```text
Start
|
+-- Need strict CI/PR control? --> continuous-pr loop
|
+-- Need task decomposition from spec? --> DAG-based loop
|
+-- Need exploratory parallel generation? --> infinite loop
|
+-- default --> sequential loop
```
## Recommended Production Stack
1. Task decomposition from spec or RFC
2. Quality gates (lint, test, type-check between iterations)
3. Evaluation harness (automated acceptance checks)
4. Session persistence (resume after interruption)
## Failure Modes
- Loop churn without measurable progress
- Repeated retries with the same root cause
- Merge queue stalls from conflicting changes
- Cost drift from unbounded model escalation
## Recovery
1. Freeze the loop
2. Audit recent iterations for root cause
3. Reduce scope to the failing unit
4. Replay with explicit acceptance criteria
No comments yet. Be the first to comment!