Test a job

Test a job after you create it and after every change: run it now, wait for it to finish, then read its log. Claude does this on its own after creating or changing a job.

Ask Claude

Run the digest now and show me the log when it's done.

Claude runs run "<job>" --wait: it starts the job now, outside its schedule, waits for the run to finish (up to 9 minutes), then prints its log, as log "<job>" does. A run still going after 9 minutes is read later with log.

What run prints without --wait

Started "Weekday open issues digest": run 7c1d9e20 issue_created. Read it with: log "Weekday open issues digest"

issue_created means Paperclip opened the run’s task for the Runner. The run itself takes from a few seconds to a few minutes.

What log prints

Job "Weekday open issues digest", run of 2026-11-03 22:45 America/Toronto: completed
Task JOB-1: done  https://your-paperclip.up.railway.app/JOB/issues/JOB-1
Summary: Open issues, 2026-11-03: 3. #1 Contact form messages are lost ...
Agent run b4259770: succeeded, 2026-11-03 22:47 America/Toronto, tokens: input 10512, cachedInput 74688, output 946
Log of agent run b4259770 (https://your-paperclip.up.railway.app/JOB/agents/<agent id>/runs/<run id>), last 30 lines (--full for all):
  [system init]
  [tool Bash]
  [user]
  [tool Skill]
  ...
  [result] The digest is done and JOB-1 is marked done.
Job: https://your-paperclip.up.railway.app/JOB/routines/<job id>

The IDs, times and numbers above are examples.

Line	What to look at
Job … completed	The run’s state. See run states.
Task … done	The task the run worked on. done is what you want; blocked comes with a reason.
Summary	The last comment on the task: the Runner’s one-line result, or the reason it’s blocked.
Agent run	Each Claude session the run took, with its tokens. A second agent run marked “follow-up of” means the first one left the task open.
Log	The end of the transcript: Claude’s messages and the tools it used.

When the test fails

• The task is blocked. Read the summary: it names what’s missing, usually a login. Redo that step on /setup. See Troubleshooting.
• The agent is paused. The log ends with “Agent Runner is paused” when Claude isn’t connected for Paperclip yet. Connect Claude on /setup.
• The result is wrong. Ask Claude to fix the skill, then test again. The Runner always runs what is pushed.

For the full transcript, ask for the log “in full” (--full), or open the run’s link in Paperclip.