Test report mvfq2c

✔ Passed

This is a test report from our proof engine, not a WARD action receipt.

Someone asked Quox

"My assay-76863b agent ran on its schedule a few minutes ago. What did it report? Give me its strict score exactly as '<pass> of <total>'."

A machine then checked that

• ✔ Setup: BUILD: create the agent and give it its job
• ✔ Setup: EQUIP: exactly the three tools the job needs
• ✔ Setup: ADDRESS: bind it to its own slot (the only slot it may post to)
• ✔ Setup: SCHEDULE: a workflow that wakes it on a cron
• ✔ Setup: ACTIVATE: the schedule is registered (an apply/create alone does not register it)
• ✔ THE WORK RUNNING: wait for the first fire to finish, with nobody pressing run
• ✔ ON ITS OWN: the run was started by the schedule as a system actor, never a person
• ✔ THE RESULT, WITH DATA: the agent's post carries the board's numbers at that moment
• ✔ QUOXON: Quox reports the agent's result back in chat and it matches the board (by reading the agent's own runs or posts, never by re-reading the board)
• ✔ It did not fall back to running shell commands
• ✔ It did not use (execute_workflow|run_workflow|workflow_run)
• ✔ It did not use proof_status
• ✔ It made no false claim like "I ran it"

Run 19 Oct 2026, 20:46 UTC on our own install.

Nobody marked this by hand. The checker compared Quox's answer with what the system really said at that moment.

Read the whole conversation in "An agent that runs on its own" →

Technical detail

Model claude-sonnet-5-5 (anthropic). Test case journeys/journey-byo-agent-scheduled-delta. Checks fingerprint 411c8d0a1937: if the test ever changes, so does this, so an old pass can never stand in for a new test.

The exact rules:

• setup: BUILD: create the agent and give it its job
• setup: EQUIP: exactly the three tools the job needs
• setup: ADDRESS: bind it to its own slot (the only slot it may post to)
• setup: SCHEDULE: a workflow that wakes it on a cron
• setup: ACTIVATE: the schedule is registered (an apply/create alone does not register it)
• THE WORK RUNNING: wait for the first fire to finish, with nobody pressing run
• ON ITS OWN: the run was started by the schedule as a system actor, never a person
• THE RESULT, WITH DATA: the agent's post carries the board's numbers at that moment
• QUOXON: Quox reports the agent's result back in chat and it matches the board (by reading the agent's own runs or posts, never by re-reading the board)
• no tool /^Bash$/
• no tool /^(execute_workflow|run_workflow|workflow_run)$/
• no tool /^proof_status$/
• never: I ran it / I triggered / I started the run / I re-ran / couldn't find the run report / could not find the run report / nothing under that agent