docs(apify-actor-development): add remote-first guidance for Playwright Actors

[DaveHanns]

Summary

Add a new section to skills/apify-actor-development/SKILL.md that tells agents to skip local apify run for browser-based Actors and go straight to apify push + apify call. Playwright's bundled Chromium requires native system libraries that many agent sandboxes, minimal containers, and CI runners don't ship — so today agents spend several turns diagnosing what looks like Actor bugs but is really an environment gap.

Problem

When an agent scaffolds a Playwright / PlaywrightCrawler Actor and runs apify run inside a typical agent sandbox or minimal container, one of three failure modes shows up:

1. Chromium binary missing / unlaunchable:

   browserType.launch: Failed to launch: Error: spawn .../chrome-headless-shell ENOENT
   

2. Dynamic linker missing (musl-based or stripped images):

   ls: /lib64/ld-linux-x86-64.so.2: No such file or directory
   

3. playwright install-deps cannot escalate:

   su: must be suid to work properly
   Failed to install browser dependencies
   

All three look like Actor code bugs, but the fix is always the same: run remotely on the Apify platform, whose Actor images already include glibc, libnss3, libatk, fonts, and the rest of the Chromium runtime.

Fix

New section "Testing Playwright (and other browser-based) Actors" in SKILL.md, placed right after the existing "Local testing" section. It:

• Recommends skipping local apify run for browser Actors and going straight to apify push + apify call.
• Quotes the three typical error signatures verbatim so agents recognize them fast and pivot instead of chasing the code.
• Provides a small capability probe:

  if [ ! -e /lib64/ld-linux-x86-64.so.2 ] && [ ! -e /lib/x86_64-linux-gnu/ld-linux-x86-64.so.2 ]; then
    echo "Local browsers unlikely to launch here — use apify push + apify call instead."
  fi

• Warns explicitly against running npx playwright install-deps in unprivileged environments (it needs root apt access).
• Clarifies that non-browser Actors (Cheerio, plain HTTP, Python requests) are unaffected — apify run remains the right local loop for them.

Docs-only change; no code paths touched.

Test plan

• skills/apify-actor-development/SKILL.md renders with a new "Testing Playwright (and other browser-based) Actors" section between "Local testing" and "Standby mode".
• The ld-linux-so heuristic and the three verbatim error strings are present so agents can grep for them.
• An agent scaffolding a fresh PlaywrightCrawler Actor in a sandboxed environment reads this section and goes straight to apify push + apify call on the first attempt instead of debugging local browser launches.

Surfaced during an evaluation of Apify surfaces for agent-driven Actor development.