chore(testing): remove the one-off browser-tools benchmark

[kyle-sexton]

No related issue: repository cleanup the owner asked for after #6477 merged.

Summary

#6477 added a browser-CLI benchmark under plugins/testing/benchmarks/browser-tools/. That directory sits inside the shipped testing plugin, so every install would carry a fixture server, CDP drivers, and raw run logs that no user runs. The benchmark was a one-off. Its verdict, playwright-cli stays the default and agent-browser is not adopted, is already recorded where it is used: ADR 0056 under Alternatives considered, plus the playwright README and ## Gotchas. This PR removes the benchmark and the clutter that came with it.

Fix

• Delete plugins/testing/benchmarks/ (52 files) and its unreleased changelog fragment.
• Add a bump: none fragment for testing. It records that the plugin ships unchanged from 0.23.0, so the release bot deletes it and releases nothing.
• .github/workflows/pr-require-checks.yml: drop the Spell-check exclusion that only covered the benchmark's results folder.
• ADR 0056: the agent-browser entry keeps the result and states that the benchmark is not kept. Its revisit trigger is now a behavior change in agent-browser.
• /playwright:playwright: the Gotchas evidence line cites ADR 0056 instead of the deleted FINDINGS.md. The plugin goes to 0.10.2, on top of the 0.10.1 docs-link release that landed on main meanwhile, with a CHANGELOG entry.

Verification

• git grep benchmarks/browser-tools returns nothing.
• Repo gates run locally on the head merged with current main all pass:
  • changelog fragments (--check, --check-required)
  • changelog parity (--check-bump, --check-order, --check-preserved)
  • changed skills, docs-only gate, lane coverage, ADR numbers
  • purged em dashes, declared prerequisites, stale-base overlap
  • plan-test-lanes (0 unmapped)
  • markdownlint and typos on the changed files

Related

• feat: benchmark browser CLIs, add a weekly upstream drift check, and tighten the playwright skill #6477 (added the benchmark)
• ADR 0056 (where the result lives)

🤖 Generated with Claude Code

https://claude.ai/code/session_01Py2vMPypYELv5S61e1sCyK

[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 2026-10-07T15:46:25.773123Z	b71e063	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.

[claude]

Claude finished @kyle-sexton's task in 44s —— View job

---

Security review (/review:security-review)

• Gather context (PR diff, changed files, custom instructions)
• Run skip gate / criteria check
• Post result

Result: no security review needed — skip gate applies.

This PR is a pure deletion + doc-pointer cleanup, with no security-relevant surface:

• 52 files deleted under plugins/testing/benchmarks/browser-tools/ (fixture HTML pages, a CDP driver, a fixture server, run scripts, and recorded run output/results). Pure removal, no new code paths.
• .github/workflows/pr-require-checks.yml (diff): only narrows the spell-check exclude: list by dropping the now-dead plugins/testing/benchmarks/*/results/** entry (the directory it exempted no longer exists). That widens scan coverage, not a weakened control.
• ADR 0056 and plugins/playwright/skills/playwright/SKILL.md Gotchas: citations repointed from the deleted FINDINGS.md to the ADR's "Alternatives considered" section. Content preserved, no guardrail removed — checked against the instruction-surface-deletion lens and this doesn't touch a CLAUDE.md/AGENTS.md/rules file/hook allow-deny list.
• Changelog fragments, plugin version bump (playwright 0.10.1 → 0.10.2): metadata only.

No injection surfaces, no auth/access-control changes, no secrets/token handling touched. Nothing to report.

[claude]

Claude finished @kyle-sexton's task in 2m 9s —— View job

---

Code review (/review:code-review)

• Gather context (PR diff vs origin/main, PR body, CLAUDE.md/AGENTS.md conventions)
• Review non-deletion files for correctness and convention alignment
• Verify no dangling references to the removed benchmark
• Verify changelog-fragment and version-bump mechanics
• Report findings

Scope. 59 files changed, +17/-8679. 52 of the deletions are the removed plugins/testing/benchmarks/browser-tools/ tree (fixtures, CDP drivers, run logs) — no review surface there. The actual review surface is 7 files: the new/removed .changes/testing/ fragments, the CI workflow's spell-check exclusion, ADR 0056, and the playwright plugin's version/changelog/SKILL.md.

Checks performed:

• git grep -n "benchmarks/browser-tools" and git grep -n "FINDINGS.md" on the HEAD tree both return nothing — no dangling path references to the deleted directory.
• The testing plugin is listed in scripts/fragment-plugins.txt, so the bump: none fragment (.changes/testing/claude-gracious-wright-1adkab-a175df31.md) is the correct release mechanic, and plugins/testing/.claude-plugin/plugin.json correctly stays at 0.23.0 (unreleased fragment added and removed in the same unreleased window, never shipped). Confirmed against scripts/lib/changelog-fragments.sh / scripts/check-changelog-fragments.sh, which require a non-empty reason line for bump: none — present here.
• playwright is not in scripts/fragment-plugins.txt, so the direct plugin.json version bump (0.10.1 → 0.10.2) + CHANGELOG.md entry is the correct mechanic for that plugin, and the two are consistent with each other.
• ADR 0056's updated "Alternatives considered" entry for agent-browser replaces the now-deleted-file pointer with a prose statement that the benchmark isn't kept, plus a new revisit trigger ("when agent-browser waits on clicks and reports uncaught page exceptions") — consistent with the benchmark's own stated finding (agent-browser's console/errors missed an uncaught exception playwright-cli caught every rep).
• /playwright:playwright's ## Gotchas evidence line now cites ADR 0056 instead of the deleted FINDINGS.md, satisfying the "name a plugin by /plugin:skill, point at a doc by its live location" pattern.
• The CI workflow's spell-check exclude glob drops only the benchmark-results clause; the remaining two exclusions (vendor/**, judge-calibration cases) are untouched and the YAML remains well-formed.

One pre-existing line, plugins/playwright/README.md:82 ("adds gotchas observed in the marketplace's browser-tools benchmark"), still references the removed benchmark by name, but that file isn't part of this diff and the line carries no path/link, so it isn't a broken reference — out of scope for this review (pre-existing content on an untouched line).

No findings. The deletion is self-contained, the release/changelog mechanics for both affected plugins are used correctly, and the cross-references that needed updating (ADR 0056, playwright SKILL.md) were updated correctly.

[github-actions]

claude-security-review has reviewed this pull request through b71e063; a later push is reviewed from there.

[github-actions]

claude-review has reviewed this pull request through b71e063; a later push is reviewed from there.

[gitguardian]

⚠️ GitGuardian has uncovered 1 secret following the scan of your pull request.

Please consider investigating the findings and remediating the incidents. Failure to do so may lead to compromising the associated services or software components.

🔎 Detected hardcoded secret in your pull request

GitGuardian id	GitGuardian status	Secret	Commit	Filename
37941187	Triggered	Username Password	6428664	plugins/testing/benchmarks/browser-tools/lib/cases.mjs	View secret

🛠 Guidelines to remediate hardcoded secrets

1. Understand the implications of revoking this secret by investigating where it is used in your code.
2. Replace and store your secret safely. Learn here the best practices.
3. Revoke and rotate this secret.
4. If possible, rewrite git history. Rewriting git history is not a trivial act. You might completely break other contributing developers' workflow and you risk accidentally deleting legitimate data.

To avoid such incidents in the future consider

• following these best practices for managing and storing secrets including API keys and other credentials
• install secret detection on pre-commit to catch secret before it leaves your machine and ease remediation.

---

^{🦉 GitGuardian detects secrets in your source code to help developers and security teams secure the modern development process. You are seeing this because you or someone else with access to this repository has authorized GitGuardian to scan your pull request.}