docs: nudge agents to read matching domain skills

[MinJaeLee1]

Summary

• Add the requested one-sentence prompt to SKILL.md so agents that load the main skill consult the matching site-specific domain skill directory before inventing an approach.
• Keep the instruction gated on BH_DOMAIN_SKILLS=1, matching the existing opt-in behavior.

Closes #155

Verification

• Confirmed agent-workspace/domain-skills/ currently has 94 site directories; largest single site directory has 4 files, so this reads only the relevant site's bounded files.
• python3 markdown/frontmatter sanity check for SKILL.md
• PYTHONDONTWRITEBYTECODE=1 python -m pytest -q -p no:cacheprovider tests → 93 passed
• python -m build --outdir /tmp/browser-harness-pr-dist . → wheel and sdist built successfully

---

Summary by cubic

Add a one-line instruction to SKILL.md directing agents that when BH_DOMAIN_SKILLS=1 and a task is site-specific, they should read every file in the matching agent-workspace/domain-skills/<site>/ directory before choosing an approach. Keeps domain skills opt-in and nudges use of per-site playbooks.

^{Written for commit 226876d. Summary will update on new commits.}

[browser-harness-review]

✅ Skill review passed

Reviewed 1 file(s) — no findings.

[cubic-dev-ai]

No issues found across 1 file