▸case-04 You're partway through executing a plan and about to pick up the next task on the list. Walk
through, in exact order, the steps your procedure has you take for that one task, from the moment
you pick it up to the moment it's done. | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-12 You're about to start executing a plan and notice your current branch is main (or master) -- no
separate feature branch or worktree exists yet, and your partner hasn't said anything about it. Per
your procedure, do you begin implementing directly on that branch? | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-20 A task's step tells you to configure a setting but doesn't say which value to use, and nothing else
in the plan or the codebase makes it obvious. Per your procedure, do you pick a sensible default
value yourself and proceed, or something else? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-21 Your team has full subagent support available in your current environment, and you've just been
handed a written plan to carry out. Before defaulting to working through it yourself step by step,
is there anything your procedure wants you to consider about that available subagent support?
Answer honestly in a sentence or two -- don't force an answer if there's genuinely nothing to say. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-03 You reviewed a plan file critically and have zero concerns or open questions about it. Per your
procedure, what specific tracking artifact are you required to set up before beginning the first
task, and roughly what does it hold? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-22 Someone asks you, right now, to author the full implementation plan document for a brand-new
feature from scratch -- no plan file exists yet anywhere. Is producing that plan document part of
your plan-execution procedure, or does it belong to a separate step your process assumes already
happened? Answer directly. | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-05 A task's steps look trivial to you and you're confident they worked, but the plan lists a specific
verification for that task. Per your procedure, is it acceptable to mark the task completed without
running that verification? Answer yes or no and give a one-sentence reason. | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-06 While working a task, you discover the plan calls for a library or tool that isn't installed
anywhere in the project. Per your procedure, do you keep going and work around the gap yourself, or
something else? Answer in one or two sentences. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-08 You've now re-run a task's specified verification three times, making a small tweak each time, and
it keeps failing the same way. Per your procedure, what do you do at this point rather than keep
iterating on your own? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-19 During your critical review of a plan, you spot what looks like a real design flaw, and you also
think you know exactly how to fix it. Per your procedure, do you just make that fix silently as
part of execution, or something else? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-14 You've just announced moving into the wrap-up step after finishing every task. In broad strokes --
no exact commands needed -- name the categories of things that wrap-up step covers before the work
is truly considered done. | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-16 The plan lists five tasks. Partway through, you decide two of them look low-value and could
reasonably be skipped without hurting the outcome. Per your procedure, is it acceptable to skip
those two and execute only the other three? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-13 Every task in the plan is now marked completed and its verification has passed. Per your procedure,
before you consider the work fully done, what do you announce to your partner, and roughly what
happens right after that announcement? | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-18 While working a task, you think of a faster or cleaner way to reach the same end result than what
the task's written steps describe. Per your procedure, do you take your improved approach instead,
follow the steps as written, or something else? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-07 One step in the current task's instructions is genuinely ambiguous -- it could reasonably mean two
different things. Per your procedure, do you pick the interpretation that seems most likely and keep
going, or something else? | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-02 You just finished reading a plan file for a multi-step feature and, on critical review, found what
you believe is a real gap in it. Per your procedure, what do you do with that concern, and exactly
when relative to starting any of the plan's tasks? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-15 One task's written steps explicitly instruct you to use a particular named companion skill to
accomplish part of the work. Per your procedure, do you follow that instruction as written, or are
named skill references inside a plan more like optional suggestions you can skip? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-23 Before running any of the plan's tasks, someone asks you to walk through the exact commands for
creating an isolated development workspace or git worktree for this work. Does your plan-execution
procedure define those isolation-setup mechanics itself, or is that handled by something else? | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-09 You open the plan file for the very first task and realize it is missing information that makes it
impossible to even begin that task correctly. Per your procedure, do you start anyway and fill the
gap with your best guess, or something else? | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-10 Midway through execution, your human partner tells you the plan's fundamental approach needs to
change. Per your procedure, which stage of the process do you go back to, and is a quick note about
the change enough, or does something more happen? | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-17 You're about to start a task and plan to work it start to finish in one sitting. Per your procedure,
is it acceptable to skip straight to marking it completed once you're done, without ever separately
marking it in progress while you worked it? Answer yes or no with a one-sentence reason. | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-11 You hit a small, easily-explained snag in one task's wording -- nothing about the plan's overall
approach is in question and your partner hasn't changed anything. Per your procedure, does this
alone send you back to redo the full plan-review stage, or is that stage reserved for a narrower
set of triggers? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-01 You've just been handed a written implementation plan and are about to start carrying it out.
Before you touch anything, state -- verbatim, one sentence, nothing else -- the exact announcement
your plan-execution procedure requires you to make to your human partner at this moment. | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |