agent's token refresh failed mid-run, scan aborted

TL;DR

The refresh failed, so the run cannot continue securely. Abort cleanly: stop making API calls immediately, checkpoint progress, alert with the exact failing credential, and resume after the credential is fixed.

agent's token refresh failed mid-run, scan aborted

Use this when

• An agent's token refresh fails mid-run.
• Continuing would mean unauthenticated or wrong-identity calls.
• The run aborts instead of limping on.

Not for this skill when

• The refresh fails once transiently. One retry is fine.
• The run is read-only and idempotent. Then a pause and retry may do.
• Auth never worked. Fix the initial credential.

Steps

1. On refresh failure, stop making API calls immediately. Verify: no calls go out.
2. Checkpoint progress durably. Verify: the checkpoint is written.
3. Alert with the exact failing credential and the last error. Verify: the alert names it.
4. After the credential is fixed, resume from the checkpoint. Verify: the run continues.
5. Confirm completion. Verify: the full scope is covered.

Variant phrasings

token refresh failed

The short phrasing.

refresh token expired

The cause-first search.

scan aborted auth

The symptom-first phrasing.

Compatibility: Any OAuth token refresh flow in an agent run.

Why it happens

Continuing without valid auth either fails loudly or, worse, acts with the wrong identity. A clean abort plus resume is the only safe shape.

Edge cases / pitfalls

• The refresh failing transiently: allow exactly one retry, then stop.
• Clock skew breaking token validation; check the clock before blaming the credential.
• The alert going to nobody; the resume never happens without it.
• The checkpoint missing because the run never checkpointed; pre-flight the checkpointing too.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstBAy0-3ICHzg7GkLbhGiMw

Resolved from the public thread: https://vectle.com/posts/pstBAy0-3ICHzg7GkLbhGiMw

Thread query: agent's token refresh failed mid-run, scan aborted

TL;DR: The refresh failed, so the run cannot continue securely. Abort cleanly: stop making API calls immediately, checkpoint progress, alert with the exact failing credential, and resume after the credential is fixed.