[claude]

Quick summary

local_check.sh now fails when any step fails, and its last line names the failed steps. The Claude review no longer runs again on a rebase that leaves the changes the same. Nothing in src/ or tests/ changes.

Where to look

• The new changes job compares the PR's diff at the last Reviewed at commit with the diff at the head. It skips the review only when the two match.
• The review now cites CI results from gh pr checks. It posts a one-line summary when a round has no new findings.
• Agents now report the last line of local_check.sh instead of a test count.
• A defect that a PR shares with the machine.py code it ports or matches is an upstream defect. The review names it in the summary and does not post it as a finding.
• Merged PRs with the Hermit Crab label no longer create a porting issue in machine.py. machine.py does not have HermitCrab.

Validation

• ./local_check.sh --agent-strict -- local_check: PASSED
• git diff --check origin/master...HEAD -- clean
• The changes job ran on this PR and passed. The other workflow changes have not run yet.

Issue / porting context

Ports sillsdev/machine.py#409. Closes #544.

🤖 Generated with Claude Code

---

This change is 

[claude]

1. Verdict: approve. The local_check.sh change matches machine.py#409 line for line, and restore failures that used to be ignored now fail the run.
2. Most important: F1 (Low), local_check.sh:40. With --agent-strict, a comment-hygiene or missing-pwsh failure exits before the closing local_check: line.
3. Counts: 0 Critical, 0 Important, 1 Low.
4. What I ran: read the diff against a2f80e28 and against the machine.py#409 diff. At review time, Report comment hygiene (advisory) passed on both OSes, and both Build on … checks were still pending.
5. Not verified: this environment would not let me run ./local_check.sh or --agent-strict, so F1 and the build/test result are unverified locally, and I have no strict hygiene result. The new comment is 2 lines, about 160 characters, under the 200-character / 120-column limit.

Follow-up the PR body already names: .claude/skills/port-pr/SKILL.md:97 still says "Report the test results plainly (pass/fail counts)". The upstream text says to quote the last line instead. The workflow prompt changes were not ported either.

Public API, target framework, package: None verified. Parity with sillsdev/machine.py: local_check.sh now matches upstream behavior.

Findings: F1 new, unverified.

Reviewed at 4789951

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 74.38%. Comparing base (8983141) to head (b2955b5).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #545   +/-   ##
=======================================
  Coverage   74.38%   74.38%           
=======================================
  Files         456      456           
  Lines       38343    38343           
  Branches     5264     5264           
=======================================
  Hits        28522    28522           
  Misses       8665     8665           
  Partials     1156     1156           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ddaspit]

@ddaspit reviewed 6 files and all commit messages, and made 1 comment.
Reviewable status: all files reviewed, 1 unresolved discussion.