The question, in the user's words: "I am changing how JpaUserCredentialStore.createCredentialEntity sets a new credential's priority. Which existing tests exercise it, and what is left untested?"

What you get at the end: the tests that already reach the change, how far up each one sits, the production paths with no test on them, and the places to add one.

The steps

1. Every caller three levels up, as places, with tests marked

niro ask get_reverse_call_chain method_fqn=org.keycloak.models.jpa.JpaUserCredentialStore.createCredentialEntity depth=3 answer=locations

JpaUserCredentialStore.createCredentialEntity has 2 direct callers and 12 more further up (14 in all).
confidence: complete | has_more: false — this is the COMPLETE caller chain. Use it directly.
model/jpa/.../JpaUserCredentialStore.java:80  ...JpaUserCredentialStore.createCredential_Nzv6uN (depth 1)
model/storage/.../UserCredentialManager.java:122  ...UserCredentialManager.createStoredCredential_Nj44zN (depth 2)
model/infinispan/.../SubjectCredentialManagerCacheAdapter.java:53  ...createStoredCredential_N778tN (depth 3)
services/.../PasswordCredentialProvider.java:133  ...PasswordCredentialProvider.createCredential_N0xesN (depth 3)
tests/base/.../CredentialModelTest.java:56  ...CredentialModelTest.testCredentialCRUD_NkjyaN [test] (depth 3)
tests/base/.../CredentialModelTest.java:176  ...testCredentialUpdateWithDuplicateLabel_NlanoN [test] (depth 3)
testsuite/.../LDAPUserMultipleCredentialTest.java:100  ...afterImportTestRealm_NjqfxN [test] (depth 3)
testsuite/.../UserStorageTest.java:938  ...UserStorageTest.testCredentialCRUD_Ni2geN [test] (depth 3)
testsuite/.../UserStorageTest.java:1042  ...testCRUDCredentialsOfDifferentUser_Ns8n5N [test] (depth 3)
testsuite/.../RecoveryAuthnCodesAuthenticatorTest.java:329  ...createRecoveryAuthnCodesForUser_Nyz8kN [test] (depth 3)
model/jpa/.../JpaUserProvider.java:1065  ...JpaUserProvider.createCredential_N0sf7N (depth 1)
[shown all 14 callers; ask without answer= for signatures and code]

[test] marks a caller in test code. Six tests can reach the change in the code as written, all three levels up, through a user's credential manager.

Check the branches that choose a class at run time. Two of the six, both in UserStorageTest, use a user stored outside Keycloak's own database (a federated user, here "thor"). UserCredentialManager.getStoreForUser sends such a user's credentials to a different store, so at run time those two tests never reach createCredentialEntity. Four tests really run it. Niro lists every caller the code makes possible; which one a running program takes can depend on data, and only reading that branch tells you.

Do not combine this with purpose=VERIFY. VERIFY gives names only and leaves the [test] marks out.

2. What production paths have no test?

The same list shows four credential providers (OTPCredentialProvider, PasswordCredentialProvider, RecoveryAuthnCodesCredentialProvider, WebAuthnCredentialProvider) at depth 3 with no test on the list beside them. Ask one level further for each, or ask the provider's own callers, to see whether a test reaches it from higher up:

niro ask get_reverse_call_chain method_fqn=org.keycloak.credential.PasswordCredentialProvider.createCredential depth=2 answer=locations

3. Write the test next to an existing one

CredentialModelTest is the closest model test: it calls user.credentialManager().createStoredCredential(...) inside a session. Open it with get_source_code at the line the list gave.

Why the answer is complete, and where it is not

What grep would do instead

To get here by search, an agent needs to know that createCredentialEntity is reached through createCredential, then createStoredCredential, then the credential manager interface, before it can search the tests for createStoredCredential(. That is three searches and three reads before the useful one, and the middle step names an interface, not a class.

Cost

What these steps cost depends on your code and your question. How Niro counts the size of its answers, and how to measure the cost on your own tasks, is in Tokens, speed and accuracy.

Choose what to test, Playbook: change code safely, Playbook: review a change.