Find Missing or Incomplete Step Logs
Select the correct attempt, inspect command or agent activity, and distinguish missing, delayed, and truncated logs.
Use this guide when a step's output is absent, still arriving, or marked truncated.
Before you begin
You need access to the run and the job or step whose output you expect.
Select the exact attempt
Open the run and select the correct run attempt. Select the job execution when a listening job ran more than once. Select the step attempt when a feedback loop restarted it.
Many apparent log gaps come from viewing an older execution or attempt.
Inspect the stream
Run steps show standard output and standard error. Agent steps show messages, tool calls, tool results, token use, cost, and the final response.
Start with the step state:
- Pending or skipped: the step has not produced a log stream.
- Running: output may still arrive while the runner drains new chunks.
- Finished: the stored stream should represent what Shipfox received for that attempt.
For an agent step, open tool calls as well as the final response. A short final message can still follow a long tool session.
If logs are missing
Confirm that the step actually started. Check the runner and job state when it did not.
If the step finished but shows no expected command output:
- Confirm that the command writes to standard output or standard error.
- Select every step attempt to find output from an earlier retry.
- Compare the saved workflow source with the command you expected.
- Check whether the runner stopped before it could drain the stream.
Use Diagnose a run when the missing stream is one symptom of a wider failure.
If logs are delayed
Check whether the step is still running and whether new chunks continue to arrive. Long-running commands may buffer their own output before writing it.
If other steps on the same runner also stop updating, inspect runner health. Do not rerun only to refresh the page; first decide whether the active work is still making progress.
If logs are truncated
Shipfox marks a stream truncated when it exceeds its budget or the runner cannot finish draining it. The visible stream is incomplete, but truncation alone does not change the step result.
Reduce noisy output or split it into useful groups before the next run. Exact budgets and retention live in Limits. Use Group command logs to make a long stream easier to scan.
Confirm the cause
Record the run attempt, job execution, and step attempt you inspected. State whether the output was in another attempt, still arriving, marked truncated, or lost when the runner stopped. That result determines whether to change the command, wait, reduce output, or repair the runner.