Repository navigation
fix(sandcastle): Pin the structured-output roles and scope the review:claude rule - #445
Merged
Merged
Conversation
Contributor
There was a problem hiding this comment.
Claude Review
Overview
This PR adds a test that pins six ROLE_AGENTS roles to maxIterations === 1. It also lets CLAUDE.md make an exception to its "don't add review:claude yourself" rule. I found no blocking problems, but I could only partly check the test's inputs (see Advisory).
Changes
| File | What changed |
|---|---|
.sandcastle/lib/repo-config.test.mts |
Imports ROLE_AGENTS and adds a ROLE_AGENTS suite. For intake, planner, critique, reviewer, gate-fixer and follow-up, one generated test each asserts maxIterations is 1. |
CLAUDE.md |
The sentence "don't add it yourself" now ends "except where docs/PROCESS.md says to (a push that needs a new review, or one more review past the cap)". |
Advisory (not blocking)
.sandcastle/lib/repo-config.mtswasn't in the context pieces, so I couldn't confirm it exportsROLE_AGENTS, keys it by those six role names, and gives each entry a numericmaxIterations. A mismatch would failpnpm run check:sandcastle(type-check or test), and the local gate catches that.CLAUDE.mdnow points atdocs/PROCESS.mdfor when to addreview:claudeby hand. I couldn't readdocs/PROCESS.md, so confirm it describes both cases (a push that needs a new review, and one more review past the cap)..sandcastle/check.shsaysmain.test.mtsreadsCLAUDE.md. If that test matches the old wording of the label sentence, it would fail. I couldn't read it to check.- Each role is checked in its own
it, and a role missing fromROLE_AGENTSwould fail with aTypeErrorrather than a clear message. This is a minor readability point.
No findings.
Contributor
mpaulosky
added a commit
that referenced
this pull request
Oct 26, 2026
Automated release blog posts for #445, 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>
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #445 +/- ##
=======================================
Coverage 96.65% 96.65%
=======================================
Files 42 42
Lines 568 568
Branches 100 100
=======================================
Hits 549 549
Misses 5 5
Partials 14 14 🚀 New features to boost your workflow:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
Closes #440, the two Advisory findings the local review left on the Sandcastle migration (#441).
What changed
.sandcastle/lib/repo-config.test.mtspinsmaxIterations1 forintake,planner,critique,reviewer,gate-fixerandfollow-up(a mutation offollow-upto 2 fails it).intake,plannerandcritiqueuse structured output (Output.object); the other three are single-shot by design, and the test's comment says so. The other roles' values stay free to move.CLAUDE.mdno longer contradictsdocs/PROCESS.md: "don't addreview:claudeyourself, except wheredocs/PROCESS.mdsays to (a push that needs a new review, or one more review past the cap)".Verification
scripts/gate.shpasses./pre-push-review: 2 rounds; one Advisory finding (the test comment claimed structured output for all six roles), fixed; nothing open.CLAUDE.mdhas the same sentence; it is not changed here.🤖 Generated with Claude Code