[tannevaled]

#75 removed the repeats by hand. ⛔ The fleet already has the command, and
I had not run it: landingmigrate repeats LAYOUT — "Find (and with --write
hide) what a layout says that the mantra strip already says".

Run on this layout as it stood, it names exactly the five fragments a reader
had spotted, and nothing else:

  footer   hidden while the strip shows   · BSD-3-Clause
  footer   hidden while the strip shows   CGO_ENABLED=0 ·
  sentence hidden while the strip shows   Pure Go, CGO_ENABLED=0.
  sentence hidden while the strip shows   BSD-3-Clause.
  sentence hidden while the strip shows   Pure Go, cgo disabled, BSD-3-Clause.

So this is landingmigrate repeats --write run on the layout as it was before
#75, replacing the hand-written version. Two differences, both in the tool's
favour:

• ⛔ It uses $, not . The detector's own idiom is
  {{ if not (partial "mantra-on.html" $) }}, and repeats.go says why: "a
  fragment inside {{ range }} has a repository for its dot". A hand-written
  . renders correctly at a layout's top level and the detector cannot see
  it — so the gate, once it exists, would go on failing on a page that is
  already right.
• It guards the hero's tail where I had deleted it, so the sentence comes
  back if the strip is ever turned off. That is what the convention is for.

Re-running the detector on the result reports the same fragments as KEPT, says more: review and none as hideable, which is the idempotent state.

Measured on the built page: cgo disabled gone, the footer reads "Pure Go ·
every library builds for js/wasm".

The gate is one flag away, and is not in this change

repeats always exits 0, so nothing can be built on it without a shell reading
its words — which is not how this fleet writes gates. The design, with the
--check flag and the one-line CI step it allows, is on
go-fleettools/fleettools#234. That repository was leased to another session
while this was written.

Refs go-fleettools/fleettools#234.

🤖 Generated with Claude Code