[petebacondarwin]

The Update OpenAPI workflow fails when refreshing #477 because Precursor's mode property now wraps its string schema in allOf: [$ref]. The temporary CLI flag guard only follows direct references, so it reports an undefined field type and stops generation. See the failed run.

Follow single-member allOf wrappers when checking field types, preserving --detection-mode and the existing checks for actual schema changes. Add regression coverage for inline, referenced, and nested wrappers, plus retyped and conflicting schemas.

Validation:

• All 24 focused guard, generated-metadata, and renamed-flag tests pass against both the pinned and latest OpenAPI schemas.
• Full root pnpm check passes against both schemas.

[pkg-pr-new]

npm i https://pkg.pr.new/cf@512

commit: faad071

[ask-bonk]

LGTM

github run

[clanky-the-clanker]

✅ Approved — This is a small, correct fix. The guard now follows single-member allOf wrappers to find a field's type, still treats multi-member allOf as ambiguous, and has regression tests. No issues found.

_{The decision is advisory: this review is a comment and never approves, blocks or requests changes on the pull request.}

---

✅ All checks passed

📋 Code Quality: ✅ LGTM

fieldType follows single-member allOf wrappers through deref for up to 10 levels, the same limit deref uses.

• A schema with more than one allOf member gives undefined. A conflicting schema therefore still stops generation with a clear staleness message.
• When a schema has both type and allOf, the direct type wins. That matches what the guard did before this change.

The new tests cover inline, referenced and nested wrappers, a wrapped field whose type changed, and conflicting allOf members. Each test checks the exact staleness message or override result.

No changeset is needed because the change only touches the internal generator guard.

---

📝 Documentation: ✅ LGTM

No documentation changes are needed for this internal generator fix.

---

_{Reviewed commit: faad0713}

[github-actions]

cf --help startup

	mean	stddev	min	max
PR	180.2 ms	3.9 ms	173.2 ms	188.3 ms
base	183.5 ms	2.4 ms	176.8 ms	186.7 ms

Δ (PR vs base): ↓ -3.3 ms (-1.8%)

Measured with hyperfine on ubuntu-latest (3 warmup runs, 20 timed runs). CI numbers carry ±a few ms of runner jitter; treat small deltas as noise.