Skip to content

chore(sandcastle): Rename the sandbox gate to sandbox-gate.sh - #347

Merged
mpaulosky merged 3 commits into
mainfrom
chore/rename-sandbox-gate
Nov 18, 2026
Merged

mpaulosky merged 3 commits into
mainfrom
chore/rename-sandbox-gate

Conversation

@mpaulosky

Copy link
Copy Markdown
Owner

Requested by Matthew · project thread

Why

The repo-ci-baseline Template now owns .sandcastle/check.sh as Sandcastle's check for the local gate. This repo uses its own .sandcastle/check.sh as the
Docker-free sandbox gate, so apply.sh keeps the repo's copy and reports a conflict. The Template's references/sandcastle.md (steps 8 and 9) says to rename
the gate to .sandcastle/sandbox-gate.sh in a PR that merges before the next re-Apply. This is that PR, ahead of the November Baseline Rollout.

What changed

  • .sandcastle/check.sh is renamed to .sandcastle/sandbox-gate.sh, and GATE_COMMAND and GATE_SCRIPT in .sandcastle/lib/repo-config.mts name it. One comment inside the script names its new file.
  • The role prompts (review-prompt.md, roles/follow-up.md, roles/gate-fixer.md) name the new file, so Sandcastle keeps working until the re-Apply replaces them.
  • roles/shared-rules.md says that a role prompt's scripts/gate.sh means .sandcastle/sandbox-gate.sh in the sandbox, as TicketManager's does, so the Template's prompts work after the re-Apply.
  • .sandcastle/CODING_STANDARDS.md and .github/ci/gate-checks.sh (a comment and an echo) name the new file.
  • New .sandcastle/repo-scripts.test.mts pins the gate's name, that its file is committed, that CHECK_PATHS covers it, and the shared-rules.md note.
    It doesn't require check.sh to be absent, since the re-Apply writes the Template's check.sh beside the gate.

Verification

  • apply.sh --dry-run against this branch reports no conflicts.
  • scripts/gate.sh passes (the pre-push hook ran it), including pnpm run check:sandcastle and the E2E tests.
  • The local pre-push review ran 3 rounds; the last found nothing.

Review notes

The local review raised these and they were accepted, not changed:

  • .sandcastle/lib/repo-config.mts:81: Once this merges, a Sandcastle branch or PR cut before the rename fails the host gate (no sandbox-gate.sh at its head) until main is merged into it, and is handed back as sandcastle:needs-human. Accepted: Matthew accepted 2026-11-18: no open PR or non-main branch exists in the repo, so no Sandcastle branch predates the rename; any later one fails closed until main is merged in.

Generated by Claude Code

The repo-ci-baseline Template now owns .sandcastle/check.sh as Sandcastle's
check for the local gate, and Apply keeps this repo's copy while GATE_COMMAND
and GATE_SCRIPT name it. Rename the repo's Docker-free sandbox gate to
.sandcastle/sandbox-gate.sh and point both constants and every mention at
it, so the next re-Apply can bring in the Owned check.sh.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KKGWRKuqUVhPuDDxfXLWYP
…les.md

The Owned role prompts the next re-Apply brings say to run scripts/gate.sh,
which needs Docker. Say in the repo-owned shared-rules.md that in the sandbox
it means .sandcastle/sandbox-gate.sh, describe what that gate runs, and pin
the gate's name, its file and the note in repo-scripts.test.mts.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KKGWRKuqUVhPuDDxfXLWYP
…dbox gate

repo-scripts.test.mts asserted that .sandcastle/check.sh is absent, but the
next re-Apply writes the Template's Owned check.sh once no gate constant names
it, which would turn the Sandcastle check red. Pin only that the gate script
is committed and isn't named check.sh.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KKGWRKuqUVhPuDDxfXLWYP
@mpaulosky mpaulosky self-assigned this Nov 18, 2026
@github-actions

Copy link
Copy Markdown
Contributor

Test Results Summary

880 tests   874 ✅  1m 40s ⏱️
  7 suites    6 💤
  7 files      0 ❌

Results for commit 953698f.

@mpaulosky
mpaulosky marked this pull request as ready for review November 18, 2026 20:03
@codecov

codecov Bot commented Nov 18, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

@mpaulosky mpaulosky added the review:claude Claude reviews each head, a backup to Copilot's review label Nov 18, 2026 — with Claude

@github-actions github-actions Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Claude Review

Overview

PR #347 renames the Docker-free sandbox gate from .sandcastle/check.sh to .sandcastle/sandbox-gate.sh. The script body is byte-for-byte unchanged apart from one comment. GATE_COMMAND and GATE_SCRIPT now point at the new name, and the role prompts, coding standards and gate-checks.sh messages follow. Two additions: a note in shared-rules.md that a role prompt's scripts/gate.sh means the sandbox gate inside the sandbox, and a new test file pinning the rename. I found no blocking problems, so there are no findings. I read the whole diff but could not search the rest of the repo for leftover references to check.sh.

Changes

File What changed
.sandcastle/check.sh → .sandcastle/sandbox-gate.sh Renamed, mode 100755 kept. Only the comment that named the script changed (check.sh runs the Sandcastle tests… is now sandbox-gate.sh runs…). Lints, repo checks, build, non-Docker tests and check:sandcastle are the same as before.
.sandcastle/lib/repo-config.mts GATE_COMMAND and GATE_SCRIPT now name .sandcastle/sandbox-gate.sh, and the comments were updated. CHECK_PATHS still covers the file through .sandcastle/.
.github/ci/gate-checks.sh Comment and echo text now say sandbox-gate.sh. The SANDCASTLE_CHECK logic is unchanged.
.sandcastle/repo-scripts.test.mts (new) Tests that the constants name the new file, that the file exists, that CHECK_PATHS covers it, and that shared-rules.md contains the scripts/gate.sh mapping sentence.
.sandcastle/CODING_STANDARDS.md, review-prompt.md, roles/follow-up.md, roles/gate-fixer.md Name the new script.
.sandcastle/roles/shared-rules.md Name the new script, and add a bullet mapping a role prompt's scripts/gate.sh to the sandbox gate.

Advisory (not blocking)

  • follow-up.md and gate-fixer.md are role prompts, and shared-rules.md and repo-scripts.test.mts describe such prompts as Template-Owned. The new test comment says shared-rules.md "survives the next re-Apply". If the Template owns follow-up.md, gate-fixer.md or review-prompt.md, the next Apply may put their old text back. The new shared-rules.md bullet is what keeps agents on the right script in that case, which fits the design.
  • I couldn't check which files are Owned. Look at the Template's Owned list. Any hand edit to an Owned prompt will be lost on re-Apply, and the follow-up role is told not to edit such files.
  • In repo-scripts.test.mts, assert.notEqual(GATE_SCRIPT, ".sandcastle/check.sh") repeats what the first test already proves, because that test pins GATE_SCRIPT to the new name. The test name also claims more than it checks. The comment on that test admits that only the name is pinned.
  • The same file's CHECK_PATHS test passes before and after the rename, as its own comment says. It only guards later edits to CHECK_PATHS.
  • repo-config.mts and the test file now have comment lines well over the surrounding width. This is cosmetic.
  • The accepted review note covers old branches failing the host gate until main is merged in. I did not re-raise it. Nothing in the diff changes that reasoning.
  • I could not search for other .sandcastle/check.sh references, such as lib/*.mts code or tests, a Dockerfile or docs. The context pieces show none. The PR author's own tests should catch a missed reference.

No findings.

@mpaulosky
mpaulosky merged commit de66a67 into main Nov 18, 2026
39 checks passed
@mpaulosky
mpaulosky deleted the chore/rename-sandbox-gate branch November 18, 2026 20:24
mpaulosky added a commit that referenced this pull request Nov 18, 2026
Automated release blog posts for #347, opened by the release workflow.
It holds every Release whose post isn't on main yet, rebuilt from main
on each run. The [skip-release] title marker keeps its merge from
starting another release.

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

review:claude Claude reviews each head, a backup to Copilot's review

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants