PR review report template

Write tmp/pr/pr-<number>.md using this exact section order. Replace every placeholder and remove any instruction that does not apply. Keep the report short.

PR #<number>: <title>

PR: <direct GitHub URL>

Base: <full SHA>

Head: <full SHA>

Reviewed: <ISO-8601 timestamp>

Problem

Trying to solve: <plain description of the actual behavior>
Linked issue(s): <direct URL(s), or “None linked”>
Before the PR: <verified base behavior>
Expected behavior: <verified contract or intended behavior>
Sources: <direct issue/PR/documentation links and source paths at immutable SHAs>

No linked issue is not a concern by itself.

Verify

Research: <what the issue, code, comments, and official documentation establish>
Isolation: <base/head fixture locations, runtime, and relevant environment>
Base result: <exact check and observed result>
PR result: <the identical check and observed result>
Regression checks: <relevant commands and their observed results>
Finding: <whether the PR fixes the stated problem and preserves existing behavior>

Include commands only if they were actually run. Quote decisive output when useful.

Risks

Scenario tied to changed code	Why it applies	Check and observed result	Likelihood
<specific supported usage or failure path>	<changed path/selector/API and real consumer>	<focused check and result>	<Confirmed / Possible / Ruled out>

Use only:

Confirmed when a failure was reproduced.
Possible when a real supported path and failure mechanism exist but a necessary check could not resolve it.
Ruled out when the focused check passed.

Do not list generic, theoretical, or unknown future risks. If no Confirmed or Possible item remains, write: No concerns found in the affected scope.

Compare

Before:

text

<actual usage, configuration, source output, or observable behavior>

After:

text

<actual usage, configuration, source output, or observable behavior>

Developer experience: <one direct explanation of what changes for users or maintainers>

Result

Safe to merge: <YES or NO>.

Confidence: <0, 50, or 100>%.

Concerns: <None, or a short list of the specific Confirmed or Possible concerns.>

Use 100% with YES when the problem and fix are verified, relevant checks pass, and no concern remains. Use 50% with NO when an unresolved Possible concern remains but no failure was reproduced. Use 0% with NO when a Confirmed concern exists, the claimed fix fails, or a relevant regression is reproduced. Explain every value below 100% under Concerns. Do not use other percentages, scoring formulas, evidence ledgers, decision matrices, administrative policy deductions, or caveats about unknowable future bugs.