fix: add explicit UTF-8 encoding to Path.read_text()/write_text() (P1 — silent data corruption on Windows)

[AlexFucuson9]

Summary

Add explicit encoding="utf-8" to all Path.read_text() and Path.write_text() calls that omit it.

Problem

Path.read_text() and Path.write_text() without the encoding parameter default to the system locale encoding. On Windows this is typically cp1252, which silently corrupts non-ASCII content (accented characters, CJK, emoji, etc.) when the file actually contains UTF-8 data.

This is NOT caught by ruff rule PLW1514 (which only flags bare open() calls) and is a common source of cross-platform bugs.

Changes

58 files fixed across the entire codebase:

• Path.read_text() -> Path.read_text(encoding="utf-8") (21 files)
• Path.write_text(...) -> Path.write_text(..., encoding="utf-8") (58 files total)

Key areas:

• gateway/run.py (9 read_text calls)
• tools/skills_hub.py (read + write)
• hermes_cli/ (auth, main, profiles, service_manager, etc.)
• plugins/platforms/ (discord, telegram, slack, whatsapp, feishu)
• agent/ (auxiliary_client, copilot_acp_client, curator)

Impact

• P1: Silent data corruption on Windows for any file containing non-ASCII text
• JSON config files, session data, skill definitions, and logs are all affected
• UTF-8 is the correct encoding per RFC 8259 (JSON) and W3C standards

Testing

All 58 changed files pass python3 -m py_compile verification.

[tonydwb]

Code Review Summary

Verdict: Comment (LGTM)

P1 critical fix: adds explicit encoding="utf-8" to all Path.read_text() and Path.write_text() calls across 58 files. This prevents silent data corruption on Windows where the default encoding is platform-dependent (often cp1252). The fix is mechanical and thorough — every instance was updated, not just the ones that happened to fail on Windows in testing.

What looks good:

• Comprehensive sweep covering all Path I/O in the codebase
• Backwards compatible (utf-8 is what the files already are in practice)
• Prevents a real silent-corruption class of bugs

[teknium1]

Closing as resolved by PR #71078 (merged, commit d372fda): the class-wide close-out salvaged your #50655/#54241/#56385/#66856/#65440 series as the backbone (authorship preserved in git log) and swept the remaining sites, so every read_text/write_text call this PR touches is now guarded on current main — verified per-site. A CI linter rule in check-windows-footguns.py plus the AST guard test now prevent regressions. Your overlapping/split variants of the same series are being closed together; the credit for the class rests on your commits.