[kyle-sexton]

No related issue: hub L261 add-on. PR Lane's completion check for #984 reads distinct-bytes= from the review split notice and saw none in six hours of reviews (03:5x-09:0xZ on 2026-11-04).

Summary

The pr-review-cli chunks notice printed distinct-bytes= only on #984's one-review branch, where the diff fits once its copies are left out. A split review, such as melodic-software/claude-code-plugins#13097's 40+40+3 run (41780888786), printed chunks= bytes= files= excluded= and no distinct bytes. So the field was missing whenever that branch did not fire.

Fix

• .github/workflows/pr-review-cli.yml, all three job copies:
  • distinct_bytes defaults to $SCOPE_BYTES before the copy check. With no copies left out, the distinct diff is the whole scope.
  • The split notice now prints distinct-bytes=$distinct_bytes between bytes= and files=.

Verification

• pr-review-cli.test.cjs: 76/76. The split test asserts the notice line chunks=2 bytes=700000 distinct-bytes=700000 files=2,1 excluded=0, and it fails on main's workflow.
• biome: no fixes. No trailing whitespace.
• Left out: a live split review showing the field. The next review over 40 files shows it.

Related

• fix(pr-review-cli): review the diff without its copies when only it fits #984, fix(review): name binary files the review cannot read as text #985

🤖 Generated with Claude Code

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-04T09:11:40.092387Z	7333e16	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[github-actions]

claude-security-review reviewed 033db05be34e82120eed50ce24037ca6cb8919ab: 0 finding(s), 0 as inline comments.

No security findings in the incremental change since 7333e16, which covers .github/scripts/pr-review-cli.test.cjs and .github/workflows/pr-review-cli.yml. I read .claude-lane/incremental.diff and judged the change from the code alone.

• Workflow change: In the split step, which appears in three places in pr-review-cli.yml, the awk-measured size of the copy-free diff now goes into a new measured_bytes variable. distinct_bytes keeps the whole scope's size ($SCOPE_BYTES) unless copies are actually left out, and only then takes the measured value.
• What it affects: This only changes the distinct-bytes number that the step reports (its distinct-bytes output and the ::notice text). The chunking decisions, the COPIES validation (^[1-9][0-9]{0,5}$) and the identical.txt regular-file and non-symlink guard are unchanged. The value comes from wc -c on a file the step wrote, so it is numeric and not attacker-controlled text.
• Test change: The test adds an assertion that a mismatched COPIES count reports distinct-bytes equal to the whole scope. It adds no security surface.
• Trust controls: No trust control is touched. The change does not widen who is trusted, a token's scope, what untrusted text reaches a lane, or any kill switch or merge hold.
• Other surfaces: No manifest, instruction-surface or userConfig changes are in this increment. I did not consult the instruction exception register, because no standing instruction was deleted or softened.

critical-findings: 0

[github-actions]

claude-review reviewed 033db05be34e82120eed50ce24037ca6cb8919ab: 0 finding(s), 0 as inline comments.

No findings. I reviewed only the change since 7333e16 (.claude-lane/incremental.diff), in pr-review-cli.yml and pr-review-cli.test.cjs. I read the code at head but did not run the test suite.

• Workflow fix: The step used to overwrite distinct_bytes with the measured size of the copy-free diff before it knew the copy count matched. A mismatched COPIES left the final split notice reporting the measured remainder, although nothing was left out. The measured value now sits in measured_bytes. distinct_bytes takes it only inside the distinct_files + COPIES == scope_files branch, where input="$distinct" and excluded=$COPIES. The one-review branch prints measured_bytes, which is the size of the file it reviews. All three job copies (around lines 1537, 3655 and 5745) change identically.
• Test: The new mismatched-copies assertion fits the fixture. With 45 files and 11 listed as identical, the remainder is 34 files, and 34 + 9 ≠ 45, so the mismatch path runs. The 45 files are over the 40-file cap, so the diff splits and the notice carries bytes=1000 distinct-bytes=1000 excluded=0, which the added regex matches.
• Closing issue: The PR says "No related issue", so there are no acceptance criteria to check.
• Pre-existing, unchanged here: The split path writes distinct-bytes only into the notice, not into $GITHUB_OUTPUT. The later DISTINCT_BYTES env therefore reads it only from the one-review branch. This increment doesn't touch that.