## TL;DR
Treat a configuration error as a stop, not a retry: the agent reruns the same Great Expectations checkpoint after a config error, and config errors do not heal, so classify the failure and halt the loop until the configuration changes. Retries are for transient failures. A checkpoint that cannot load its configuration fails before touching data, identically, forever, and each rerun costs a full agent cycle.

```text
agent loop detected: data quality agent retries same great expectations checkpoint after config error
```

## Use this when
- The check or run fails with this symptom: an agent loop retries the same Great Expectations checkpoint after a configuration error.
- You are working on an agent workflow running Great Expectations checkpoints with automatic retries and the failure repeats on the same data or configuration.
- You need the failing rows, rule, or stage isolated before changing thresholds or code.

## Not for this skill when
- A checkpoint failing on a transient warehouse connection, where a bounded retry is the right behaviour.
- The error text names a different tool, stage, or root cause. Match the verbatim error first and use that skill.

## Steps
1. Read the checkpoint failure and classify it: configuration load, connection, or validation. Verify: the failure is a configuration error raised before validation
2. Fix the configuration fault named in the error, such as a missing store or a bad reference. Verify: the checkpoint loads its configuration in a dry run
3. Set the retry policy to retry only connection and transient classes, with a small cap. Verify: a config error now stops the workflow on first occurrence
4. Key retries on a hash of the configuration so any edit re-enables a run. Verify: an edited config runs without manual reset
5. Report the halted state with the config error attached. Verify: a human sees the blocker immediately

## Variant phrasings
### agent retries great expectations checkpoint config error loop
The shorter phrasing focused on the named failure.
### agent loop same checkpoint configuration error
The symptom phrasing a searcher tries next.
### checkpoint retry policy config error fix
The variant naming the failing stage or tool directly.

Compatibility: Great Expectations checkpoint execution, agent retry policies

## Why it happens
Retry loops usually key on exit status alone. Configuration errors and validation failures share a non-zero exit with transient faults, so an undiscriminating loop retries permanent failures and never escalates the one thing a human must change.

## Edge cases / pitfalls
- Distinguish validation failure from execution failure in the report; a failed expectation is a result, not an error to retry.
- A config hash must include referenced files such as expectation suites, not only the checkpoint file itself.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_oTlwIUDSg7u5zWLSZHBpyQ