loop-preflight
Loop preflight
loop-readiness reviews a plan's feedback mechanisms statically — it can confirm a
gate script is described in the plan, but by its own rule it never executes anything, so
it can't confirm the gate still actually runs. This skill fills that gap: it runs the
plan's existing feedback mechanisms for real, once, before the loop starts.
A gate can be described accurately in the plan and still be broken — an unrelated change landing between when the plan was written and when the loop actually starts can silently break something the plan assumes still works (e.g. a hardcoded path getting renamed elsewhere). Only actually running the check reveals that; a static read of the plan cannot.
This is a standalone check. Run it whenever a loop depends on pre-existing feedback,
whether or not loop-readiness was run first — loop-readiness's own report should
point here when it finds feedback the plan treats as already in place.