[Upd4ting]

Why

Goal: make the CI and the verify-template for this template as fast as is reasonable, without losing a check. Done means every fix is measured and in its own PR against main with before and after numbers, the checks are green, and the review gate has passed.

Every journey scaffolded the module on its own. Each one ran ajs module init with its install, then pnpm install. tooling-configs 0.3.5 adds context.prepared. It builds a folder once per run and gives each journey its own copy. This PR takes 0.3.5 and uses context.prepared for that scaffold.

What changed

• @antelopejs/tooling-configs goes from ^0.3.4 to ^0.3.5, and antelopejs-sync rewrote the synced SKILL.md.
• scaffolded() in verify.config.ts now wraps the offline manifest step, ajs module init and pnpm install in context.prepared("my-module", ...). Every journey gets a fresh copy of the installed module.

Scope

• Journeys and their checks do not change. scaffold-shape and scaffold-docs now read a module that pnpm install has also run in. They check only files and scripts, so the extra install makes no difference to them.

Blast Radius

Only agents that verify this template are affected. If the first journey's build fails its checks, the run does not keep it, and the next journey builds again.

Verification

• Full verify (new --all with every feature), 5 interleaved rounds per side. Median CPU (user plus sys) went from 136.9 s on main (0.3.3) to 88.2 s, a 36% drop. With 0.3.5 and sync only it was 127.1 s. Median wall time went from 91.5 s to 61.9 s, and the ranges do not overlap. Every run passed.
• pnpm run checks passes 8 of 8, and map passes.
• The A/B ran before perf(build): skip the duplicate lib check in build and test compiles #29 and build(deps): take tooling-configs 0.3.4 #31 merged. After the rebase onto the new main, one more full verify passed at 68.7 s CPU and 44.2 s wall.