docs(cli): do not suggest generating test scenarios in parallel

[Yury Semikhatsky (yury-s)]

Summary

• Section 2.2 of the test generation skill says scenarios must be generated in sequence, while 2.3 claimed the loop is safe to parallelise. Concurrent npx playwright test runs in the same directory are not safe, so drop that claim.

[github-actions]

Test results for "MCP"

4 failed
❌ [chrome] › mcp/cli-webmcp.spec.ts:188 › webmcp-call disambiguates same-name tools in identical same-origin frames @mcp-macos-latest-chrome
❌ [firefox] › mcp/cli-keyboard.spec.ts:19 › press @mcp-windows-latest-firefox
❌ [firefox] › mcp/cli-session.spec.ts:54 › idle timeout shuts the session down @mcp-windows-latest-firefox
❌ [firefox] › mcp/http.spec.ts:145 › http transport browser sigint @mcp-windows-latest-firefox

8688 passed, 1474 skipped

---

Merge workflow run.

[github-actions]

🟢 All four failures are pre-existing flakes — this PR is clear

Hi, I'm the Playwright bot and I triaged the failing MCP jobs.

This PR changes one line of markdown in a CLI skill reference, so it can't reach any product code path. All four failing tests also fail on other SHAs in the results DB, three of them on main pushes.

Details

Pre-existing flake / infra

• [firefox] › mcp/cli-session.spec.ts:54 › idle timeout shuts the session down — the noisy one. On mcp-windows-latest-firefox it failed 21 of 53 runs (40%), across 21 distinct SHAs, and passes on every run on ubuntu (44/44) and macOS (43/43). Recent failures on unrelated SHAs: b8627f95, f14e9f5c, dc82cd77 — all main pushes. A windows-only timing race in the idle-shutdown path.

  🟩🟥🟩🟥🟥🟥🟩🟩🟩🟩🟩🟩🟩🟩🟩🟥🟩🟥🟩🟩🟥🟩🟩🟥🟩🟥

• [chrome] › mcp/cli-webmcp.spec.ts:188 › webmcp-call disambiguates same-name tools in identical same-origin frames — also failed on #42792, #42789 and #42785, and passes 14/14 on windows.

• [firefox] › mcp/cli-keyboard.spec.ts:19 › press — 6 of 427 runs fail, windows-firefox only, 0/124 on ubuntu and 0/169 on macOS. Most recently on dc82cd77, a main push.

• [firefox] › mcp/http.spec.ts:145 › http transport browser sigint — 5 of 466 runs, again windows-firefox only, e.g. 96af9544 (push) and #42663.

Every one of these is windows-firefox or a known-bimodal chrome test, and this PR touches references/test-generation.md only — no runtime code ships from it.

_{Triaged by the Playwright bot - agent run}