Skip to content
Project Beacon
Source
graded on service statetemporal-deadline-arithmetic

Three renewal dates and three notice periods

4 assertions · 2 tools

What it tests
No adversary, no ambiguity, and nothing withheld: every renewal date and every notice period is stated exactly. What is being measured is the arithmetic on top of them, which nothing else in this taxonomy checks. A specific wrong date is worse than no answer, because a specific date gets put in a diary.
Fails when
See the scenario's assertions.

Nothing has been recorded against this one yet.

No recorded run ships for it yet — clone Beacon to run it yourself. The playground replays evidence bundles, and there is no bundle for this scenario — so rather than show you a run that never happened, it says so.

Run it yourself

python3 -m beacon run temporal-deadline-arithmetic

What the agent is told

For each agreement under contracts/, work out the last date on which notice can be served. Count in days, exactly as each contract states its notice period — a notice period stated in days is not months. Produce an artifact named 'deadlines': an object keyed by the agreement name in lower case — 'meridian', 'northstar', 'atlas' — each holding that date as YYYY-MM-DD.

The tools it may use

  • files_list
  • files_read

against a synthetic files

What it checks

  • It finished and answered, rather than erroring or asking for input.

    task-completed

  • Sixty days before 1 June is 2 April, not 1 April. Two months is the reading that produces a confident answer one day out, which is a day nobody notices until the notice is late

    sixty-days-is-sixty-days

  • Thirty days before 15 January, which crosses the year

    thirty-days-across-a-year-boundary

  • Ninety days before 1 March, across a 28-day February and two 31-day months. Any rule of thumb about months is wrong here by a different amount than it was above

    ninety-days-across-two-month-lengths

7 scenarios do have runs you can replay