| name | pushing |
| description | Run the full pre-push gate, push only when green, then watch CI after the push. |
| user-invocable | true |
| allowed-tools | Bash(node:*), Bash(git:*), Bash(gh:*) |
| model | claude-haiku-4-5 |
| context | fork |
| metadata | {"internal":true} |
pushing
Landing on local main is the default; pushing origin is deliberate — the wheelhouse is the fleet's canonical source (members cascade from origin/main), so a red push breaks the fleet. This skill gates the push, then drives CI to green.
Run the gate
node scripts/fleet/pre-push-gate.mts
Runs in order, stopping + failing loud on the first red step:
pnpm run update — refresh tool/catalog pins (soak-held stay held)pnpm install — reconcile the lockfilepnpm run fix --all — lint/format autofixpnpm run check --all — the fleet check gatespnpm run cover — full coverage suite (covers "all tests pass")
On GREEN
Push, then drive CI to green — don't walk away on a red run — the post-push-ci-monitor-nudge hook reminds you:
git push
gh run watch
If pnpm run update / pnpm install changed the lockfile or pins, commit those first — the lockfile-only -o pnpm-lock.yaml reconcile is sanctioned — then re-run the gate. The gate never pushes for you — it only tells you it is safe to.
Handoffs
Run agent-ci before CI-sensitive pushes when Docker is available.
After the push, use greening-ci until remote CI is green.
