fix(skill-creator, mcp-builder): specify utf-8 file encoding and pin mcp<2

[AbhiPra24]

Summary

• Fixes skill-creator scripts crash on Windows for any skill containing non-ASCII text (missing encoding on read_text()) #1686 & mcp-builder: evaluation.py discards a completed run on Windows — write_text() without encoding hits cp1252 on the report's emoji #1669: Add explicit encoding="utf-8" to file read/write operations across skills/skill-creator/scripts/, skills/skill-creator/eval-viewer/, and skills/mcp-builder/scripts/ so Windows environments do not fail with UnicodeDecodeError / UnicodeEncodeError (cp1252) when encountering emojis or non-ASCII characters.
• Fixes mcp-builder: requirements.txt allows mcp>=2, where connections.py fails at import (streamablehttp_client renamed) #1668: Pin mcp<2 in skills/mcp-builder/scripts/requirements.txt to avoid breaking changes with streamablehttp_client.

Changes

• skills/skill-creator/scripts/aggregate_benchmark.py: specify encoding="utf-8" in open() calls.
• skills/skill-creator/scripts/generate_report.py: specify encoding="utf-8" in read_text() and write_text() calls.
• skills/skill-creator/scripts/improve_description.py: specify encoding="utf-8" in read_text() and write_text() calls.
• skills/skill-creator/scripts/quick_validate.py: specify encoding="utf-8" in read_text().
• skills/skill-creator/scripts/run_eval.py: specify encoding="utf-8" in read_text() and write_text() calls.
• skills/skill-creator/scripts/run_loop.py: specify encoding="utf-8" in read_text() and write_text() calls.
• skills/skill-creator/scripts/utils.py: specify encoding="utf-8" in read_text().
• skills/skill-creator/eval-viewer/generate_review.py: specify encoding="utf-8" across all read_text() and write_text() calls.
• skills/mcp-builder/scripts/evaluation.py: specify encoding="utf-8" in write_text().
• skills/mcp-builder/scripts/requirements.txt: pin mcp>=1.1.0,<2.

Testing

• Verified with python3 skills/skill-creator/scripts/quick_validate.py skills/mcp-builder
• Verified with python3 skills/skill-creator/scripts/quick_validate.py skills/skill-creator

cc @98zc5g5jyw-arch for review

[Hahaknight]

Tested the encoding choice on real Windows 11 against a production Chinese-language SKILL.md — one substantive gap plus a scope note:

1. Plain utf-8 misses the BOM case documented in #1686 (reporter found a real BOM'd SKILL.md in the wild). Prepending a UTF-8 BOM (EF BB BF) to a valid skill:

• read_text(encoding="utf-8") (this PR) → content starts with , so startswith('---') / lines[0].strip() != '---' checks return False → a valid skill is silently rejected as "no frontmatter". No crash, so it fails quietly — arguably worse than the current loud crash.
• read_text(encoding="utf-8-sig") → passes, and is byte-identical to utf-8 on non-BOM files (no downside).

Suggest using utf-8-sig for the SKILL.md read sites (quick_validate.py, utils.py:parse_skill_md) to match #1687's approach and the reporter's recommendation. For JSON/HTML artifact reads/writes in eval-viewer/mcp-builder, plain utf-8 is fine.

2. Scope: the mcp>=1.1.0,<2 pin in mcp-builder/requirements.txt looks unrelated to the Windows encoding fix — might be worth splitting into its own PR so the maintainers can evaluate each independently.

The eval-viewer and mcp-builder read/write sites this PR fixes are not covered by #1687, so the coverage between the two PRs is complementary — merging both (with utf-8-sig at the SKILL.md sites) would close #1686 fully.