Two plugin test suites fail on a clean main under the pinned ruff 0.16.0

[kyle-sexton]

What

A full bash scripts/run-plugin-tests.sh sweep on a clean main (cff36aa) fails
two suites. Neither failure involves the plugin's own logic — both are the suite's
ruff lint pass reporting findings under ruff 0.16.0, which is the version the
fleet pins (dotfiles/.chezmoidata/uv-tools.yaml:35 → ruff==0.16.0), so this is
not local version drift.

plugins/source-control/skills/babysit-prs/scripts/engine.test.sh — exit 1.
The unit tests all pass:

Ran 597 tests in 16.132s
OK

The exit code comes from the lint pass:

1460 |                 raise RuntimeError("association query did not return every watched PR")
TRY004 Prefer `TypeError` exception for invalid type
603 |             raise RuntimeError(f"Malformed review-thread page info for {repo}#{number}")
TRY004 Prefer `TypeError` exception for invalid type

plugins/ruff-format/hooks/ruff-format.test.sh — exit 1, on:

FAIL: E712 not reported:

The preceding cases pass, so the harness works; this one expectation does not
hold under 0.16.0.

Why this is worth a ticket rather than a quiet fix

TRY004 asks for TypeError where RuntimeError is raised. That is an exception
type change on two error paths in the babysit engine, not a formatting nit —
anything catching RuntimeError there changes behavior. It wants deciding on its
own merits: either the rule is right for these two sites and the raises change, or
the rule does not fit a "the API returned something structurally wrong" case and
the sites get a scoped noqa with the reason recorded.

The E712 expectation needs the opposite question asked: did ruff stop reporting
E712 in the shape the fixture builds, or did the fixture stop building that shape?

How this surfaced

Found while running the full sweep to verify an unrelated branch
(#1970). git diff --name-only origin/main...HEAD confirmed that branch touches
neither plugin, so these files were running main's version. Not folded into that
PR deliberately — changing another plugin's exception semantics inside a
worktree-detection change is the scope creep the repo's conventions warn about.

A third suite, plugins/claude-ops/skills/plugins/scripts/fleet-state.test.sh,
also fails on clean main with FAIL: home exclusion: $HOME with .claude is user scope, not project context. Different root cause, same sweep — worth confirming
whether CI sees it too, since a red suite on main is either a CI gap or a
local-environment dependency the suite does not declare.

[kyle-sexton]

Root cause: an upstream default change, not two unrelated findings

Both ruff failures are one cause. Ruff 0.16.0 changed its default rule
selection — the changelog's breaking-changes
section states it
"now enables a much larger set of rules by default (413, up from 59)" and that
18 opinionated E/F rules were removed from the default set, E712 among
them.

That resolves both questions this ticket asked:

• TRY004 is not a new finding in the babysit engine. TRY is newly in the
  default set. The code did not change.
• E712 — neither of the two options in the ticket. Ruff did not stop
  reporting it and the fixture did not stop building the shape; E712 left the
  default selection. Verified locally under 0.16.0: a bare ruff check reports
  All checks passed! on x = 1 / if x == True:, while --select E712 reports
  it.

Measured on this repo (no ruff config exists anywhere in it, so it lints with
whatever upstream's defaults happen to be):

rule set	findings across plugins/ + scripts/
ruff 0.16.0 bare defaults	130
the managed standards ruff policy	385

This is a rediscovery, and two open PRs already point opposite ways

• dev-env: local ruff auto-upgraded past the CI pin (0.16.0 vs ruff==0.15.22) — 82 findings on clean main, engine.test.sh fails locally #1856 is the same problem, already filed.
• fix(dev-env): resolve ruff from the declared pin instead of PATH #1871 (open) solves it by pinning the harness DOWN to CI's 0.15.22 via a
  scripts/run-ruff.sh wrapper. Built and tested; explicitly declines to mass-apply
  the newer ruff's findings.
• chore(deps): bump ruff from 0.15.22 to 0.16.1 in /.github #1953 (open, Dependabot) solves it by raising CI UP to 0.16.0. Its CI is
  red — plugin-gate and ci-status both fail, which is the 130 findings
  surfacing. It cannot land as-is.

The CI pin is Dependabot-owned (.github/dependabot.yml, package-ecosystem: pip,
directory: /.github), so raising it is a routed PR, not a hand edit.

The durable fix is upstream

The reason an upstream default change could break this repo at all is that it has
no ruff configuration — it inherits defaults it never chose. standards owns a
ruff component whose ruff.toml pins select explicitly, and only dotfiles
consumes it today. Adopting it here makes upstream default churn a non-event.
Filed separately as the tracked migration, with the 385 figure attached.

Third suite

The fleet-state.test.sh failure is a genuine defect with a different root
cause — the $HOME exclusion compares two spellings of the same directory. Split
out as #1982 and fixed in #1983.