Release CU: stage-batch5 — 2-PR low-risk batch (v0.51.123) — gzip+ETag static caching / Open in VS Code

[nesquena-hermes]

Release CU — v0.51.123 — stage-batch5 — 2-PR low-risk batch

Fifth batch from the May 24 sweep. Two well-tested contributor PRs, both pre-cleared by Opus advisor:

PR	Author	LOC	Surface	Risk
#2779	@v2psv	+346/-4	api/routes.py _serve_static()	low — strictly /static/* scope, in-process cache with (size, mtime_ns) invalidation, 11 tests
#2787	@munim	+508/-40	api/routes.py + static/ui.js + i18n	low — context-menu addition, uses existing safe_resolve traversal guard, 26 tests

Cherry-picks

d20da832 fix(static): tighten cache validators and 304 headers           (#2779)
7999d1c7 feat(workspace): add Open in VS Code action for files (#2735)   (#2787)
79a3ac37 Stamp CHANGELOG for v0.51.123

Conflict resolution + cherry-pick fix-ups

PR #2779's cherry-pick onto current master had a real merge conflict in _serve_static() because the function body was entirely rewritten. Resolution:

1. Took theirs (PR replaces the old no-store wfile.write(raw) impl with the new cached wfile.write(body) impl)
2. Caught a regex slip where the tail wfile.write(raw) was left from master; corrected to wfile.write(body) so gzipped content is actually transmitted
3. Folded missing import gzip at api/routes.py:9 into the contributor's commit via git rebase --autosquash

Pre-Opus gate

Check	Result
Python ast.parse + node -c on all touched	PASS
Merge-marker grep	clean
New regression tests (37 across 2 files)	pass in isolation
Full pytest	6333 passed, 7 skipped, 0 failed (up 37 from 6296 — matches new test count)

Opus advisor verdict

PR	Decision	Notes
#2779	SHIP	Cache key safely bounded behind two sandbox checks; parse_qs v-param logic correct including empty-string and multi-value edge cases; tests pin both header policy and (size, mtime_ns) invalidation. Memory ceiling (~3.5 MB) is fine.
#2787	SHIP	safe_resolve runs BEFORE the path is handed to Popen, with full .resolve() normalization; Popen pattern is byte-for-byte the model used by _handle_file_reveal; Docker prefix rewrite is post-validation and admin-trusted.

What's NOT in this batch

Routed to Batch 6 (Windows stack rebase) or held for review:

• feat(start.ps1): expand hermes-agent candidate paths for Windows installers #2805/docs(start.ps1+README): clarify native Windows venv path; remove misleading WSL2-venv-portability claim #2806/harden(start.ps1): TryParse HERMES_WEBUI_PORT + exit AFTER try/finally cleanup #2807/ci(windows): add native-Windows startup E2E workflow #2811 — Windows start.ps1 follow-ups, rebasing onto current master now that feat: native Windows community-guide link + start.ps1 launcher (#1952) #2783 has landed
• style(composer): clamp composer-box max-width so wide displays don't squeeze footer chips #2812 — composer max-width clamp (needs UX review with screenshots per Nathan's standing rule)
• feat: PATCH /api/mcp/servers/{name} — enable/disable toggle #2776 — MCP PATCH (new HTTP verb + CSRF)
• fix(webui): route Cursor ACP and gate reasoning effort by model #2792 — Cursor ACP (blocked on hermes-agent#30835)
• feat(runtime): wire runner route selection harness #2794, feat: add Kanban consolidated view toggle #2814, feat(i18n): add Turkish (tr) locale #2772, feat: add Auxiliary Models settings card #2680, Add scoped workspace Artifacts tab #2673 — each needs its own review path