Best possible solution:

Keep the narrow benchmark-harness compare mode and let the existing exact-head automerge/check gates decide the PR.

Do we have a high-confidence way to reproduce the issue?

Not applicable as a feature/tooling PR. The prior PR defects were source-proven in review comments and the current head addresses them; I did not run local tests because this review was read-only.

Is this the best way to solve the issue?

Yes. The benchmark script already owns saved benchmark JSON output, and adding comparison there with focused output and routing coverage is the narrow maintainable path.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against a8b5f5d5518b.

Label changes:

• add rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🌊 off-meta tidepool and patch quality is 🐚 platinum hermit.
• remove rating: 🦪 silver shellfish: Current PR rating is rating: 🐚 platinum hermit, so this older rating label is no longer current.

Label justifications:

• P3: This is a low-risk repository benchmark-harness improvement with no production runtime blast radius.
• rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🌊 off-meta tidepool and patch quality is 🐚 platinum hermit.
• status: 🚀 automerge armed: This PR is in ClawSweeper's automerge lane. Not applicable: This is a ClawSweeper bot replacement PR, so the external-contributor proof gate does not apply directly; the linked source PR supplied copied live terminal output for the central compare-mode behavior.

PR surface:

Tests +77, Other +109. Total +186 across 4 files.

What I checked:

• PR head implements saved-report comparison: The current PR head parses --compare-baseline and --compare-candidate, builds deltas from saved reports, writes --output JSON, and prints either JSON or text output. (scripts/bench-cli-startup.ts:1048, 1afa110f1bc8)
• Output regression coverage exists: The PR test creates saved baseline/candidate reports, calls the comparison helper, writes JSON output through the shared writer, and asserts the delta fields plus parent-directory creation. (test/scripts/bench-cli-startup.test.ts:32, 1afa110f1bc8)
• Previous helper-routing blocker is repaired: The tooling expectation for test/helpers/temp-dir.ts now includes both test/scripts/bench-cli-startup.test.ts and test/scripts/test-projects.test.ts, matching the new direct importers. (src/scripts/test-projects.test.ts:857, 1afa110f1bc8)
• Routing fixture matches importer behavior: The changed-target routing fixture now includes test/scripts/bench-cli-startup.test.ts as a direct shared-helper importer and expects it in the resolved target list. (test/scripts/test-projects.test.ts:1914, 1afa110f1bc8)
• Current main lacks the requested mode: Current main has --output for saved benchmark reports but no compare-baseline/candidate mode, so the PR is not obsolete on main. (scripts/bench-cli-startup.ts:905, a8b5f5d5518b)
• Source PR provenance: The linked source PR is closed unmerged because ClawSweeper could not update that branch; this PR is the credited replacement and preserves the original contributor context. (77414e48cc35)

Likely related people:

• vincentkoc: Recent GitHub path history repeatedly touches scripts/bench-cli-startup.ts and its tests, and a maintainer command armed the original source PR for automerge. (role: recent benchmark harness contributor; confidence: high; commits: 77f8b1671629, 59365959d3b0; files: scripts/bench-cli-startup.ts, test/scripts/bench-cli-startup.test.ts)
• RomneyDa: Recent commits updated temp-dir helper routing expectations in the same changed-target test surface affected by this PR. (role: recent changed-test routing contributor; confidence: medium; commits: dfd8a2220b92; files: src/scripts/test-projects.test.ts, test/scripts/test-projects.test.ts)
• hxy91819: Recent history introduced the shared temp-dir helper guidance and helper surface that the new benchmark test imports. (role: shared test-helper contributor; confidence: medium; commits: 06431fd99b9e; files: test/helpers/temp-dir.ts, test/helpers/AGENTS.md)

• 🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
• 🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
• 🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
• 🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
• 🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
• 🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
• 🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

• ClawSweeper keeps one durable marker-backed review comment per issue or PR.
• Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
• A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
• PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
• Maintainers can also comment @clawsweeper review to request a fresh review only.
• Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
• Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
• Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.