[devs-rootstrap]

Jira board reference:

• No ticket: maintenance of the repository's push checks.

---

What does this do?

Updates the model used by Pushgate's advisory AI review in .pushgate.yml from the pinned claude-sonnet-4-20250514 to the CLI alias sonnet, which resolves to the current Sonnet release (today claude-sonnet-5-5) without further bumps.

---

Why did you do this?

The Claude CLI no longer serves claude-sonnet-4-20250514. Pushgate reports "model not found", turns that into a warning, and asks for an interactive confirmation. Any push without a terminal attached (scripts, automation, some IDE integrations) is refused at that point even though every blocking check passed. Pointing the advisory review at a model the CLI still serves lets the review run again instead of failing on every push.

Who/what does this impact?

Everyone who pushes to this repository with Pushgate installed. The blocking tools (type-check, eslint, lint, lint-translations, test) and the built-in policies are unchanged; only the advisory review's model changes.

---

Layer 2: Risk Classification

risk:low: a one-line config change to a local pre-push check.

How did you test this?

• Confirmed claude --model sonnet -p ... answers (and reports claude-sonnet-5-5), and that Pushgate passes the configured model straight through as --model.

• Not exercised through the pre-push hook itself: this branch was pushed from a fresh worktree without node_modules, where husky's hooks are not installed. A reviewer with Pushgate set up can confirm by pushing any branch after merging.

• Tested on iOS

• Tested on Android

• Tested on a small device

• Tested on a real device

• Tested all flows related with this PR changes

• Tested accessibility

• Added tests

---

Notes:

The diff_size policy (max_changed_lines: 700, mode: warning) also needs an interactive confirmation when exceeded, so non-interactive pushes of larger branches are still refused. That is left as is here; worth a separate decision.

🤖 Generated with Claude Code

[github-actions]

💯 Test Coverage

Lines	Statements	Branches	Functions
	56.23% (347/617)	59.18% (174/294)	48.76% (118/242)

😎 Tests Results

Tests	Skipped	Failures	Errors	Time
139	0 💤	0 ❌	0 🔥	23.682s ⏱️

👀 Tests Details • (56%)

File	% Stmts	% Branch	% Funcs	% Lines	Uncovered Line #s
All files	56.23	59.18	48.76	56.9

report-only-changed-files is enabled. No files were changed in this commit :)

[Copilot]

🟢 Approval recommended

No active review issues remain for the advisory model update intended to prevent model-not-found failures during non-interactive pushes.

No open issues 🎉

What changed in this PR

Overall Review: ai:clean

Updates Pushgate’s advisory review model to address the reported “model not found” failure that interrupts non-interactive pushes.

Changes:

• Replaces claude-sonnet-4-20250514 with claude-sonnet-5-5.
• Leaves blocking checks and built-in policies unchanged.

No actionable issues identified. End-to-end pre-push validation remains untested, as noted in the PR.

File	Description
.pushgate.yml	Updates the Claude model used for advisory AI reviews.

_{Review effort: Balanced}