style: replace em dashes and en dashes with standard punctuation

[jongio]

Removes every em dash and en dash from tracked source, docs, and config across 99 files (368 replacements), replacing each with the punctuation the sentence actually calls for: a colon for label-then-explanation shapes, a semicolon where two independent clauses were joined, and a comma for appositives and parentheticals.

Two focused tests cover the unchanged-configuration and dry-run branches in init.go. No functional behavior change. Production edits are limited to punctuation in comments, documentation, and user-facing strings.

Replacement was script-driven for consistency, then every added line was read by hand. The manual pass caught real defects the rules introduced, including empty markdown table cells that had held a bare dash collapsing and breaking column counts, and a sentence that ended up with two colons in it.

Excluded on purpose:

1. CHANGELOG.md. Its entries mirror published GitHub Release notes and commit subjects, so editing them desyncs the file from immutable git history.
2. web/src/pages/reference/changelog/index.astro. Generated release-note markup, with the same immutability rationale.

Verified: all non-Vitest mage preflight gates pass locally with CI-compatible pnpm 9 and Go 1.26.6, including builds, lint, docs, E2E, race tests, security scanning, and vulnerability scanning. The dashboard suite passes separately with 1,436 tests across 63 files. The parallel preflight run encountered only Vitest worker startup starvation while every executed test passed.

[github-actions]

🚀 Website Preview

Your PR preview is ready!

📎 Preview URL: https://jongio.github.io/azd-app/pr/609/

This preview will be automatically cleaned up when the PR is closed.

[codecov]

Codecov Report

❌ Patch coverage is 83.33333% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 64.07%. Comparing base (1b66ec6) to head (e112f22).

Files with missing lines	Patch %	Lines
cli/src/cmd/app/commands/run.go	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #609      +/-   ##
==========================================
+ Coverage   63.87%   64.07%   +0.20%     
==========================================
  Files         241      241              
  Lines       31502    31502              
==========================================
+ Hits        20121    20185      +64     
+ Misses      10061     9989      -72     
- Partials     1320     1328       +8     

Flag	Coverage Δ
unittests	64.07% <83.33%> (+0.20%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
cli/src/cmd/app/commands/core_deps.go	62.73% <ø> (ø)
cli/src/cmd/app/commands/core_helpers.go	73.29% <ø> (ø)
cli/src/cmd/app/commands/doctor.go	87.20% <ø> (ø)
cli/src/cmd/app/commands/init.go	66.96% <100.00%> (+9.69%)	⬆️
cli/src/cmd/app/commands/logs.go	64.07% <ø> (ø)
cli/src/cmd/app/commands/mcp.go	63.90% <ø> (ø)
cli/src/cmd/app/commands/mcp_sanitize.go	100.00% <100.00%> (ø)
cli/src/cmd/app/commands/mcp_tools.go	76.32% <100.00%> (ø)
cli/src/internal/azure/standalone_logs.go	10.68% <ø> (ø)
cli/src/internal/dashboard/envinfo/envinfo.go	86.95% <ø> (ø)
... and 21 more

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[wbreza]

A clean, inert punctuation sweep, except two generated protobuf artifacts were edited instead of their .proto source.

What stood out

• Genuinely inert edits — Go changes land in comments and doc strings, TypeScript changes in comments and test descriptions. No sanitizer allowlist, regex, exec argument, or wire constant was touched.
• Test assertions untouched — Every test-file edit sits in a comment or a describe/it label, never in an expect argument, a t.Errorf, a golden fixture, or a case name used in a comparison.
• Doc tables survived — Column counts hold across the five largest doc diffs, the one cell holding a bare dash became n/a rather than collapsing, and numeric ranges read correctly.
• Generated code drift — Two derived protobuf files were hand-edited while their .proto source kept its em dash, so the next codegen run silently reverts both.

Top issues

• [F-001] Generated protobuf file edited instead of its .proto source — cli/src/gen/proto/azdapp/v1/common.pb.go:322
• [F-002] One source comment rendered with different punctuation in Go and TypeScript — cli/dashboard/src/gen/proto/azdapp/v1/common_pb.ts:508

Cross-cutting

• [F-001] Derived artifacts edited ahead of their source (process) — Spans both generated protobuf outputs plus the untouched proto/azdapp/v1/common.proto.

Improvement opportunities

• [O-001] Doubled colons where a bold label already ends in a colon (Optional) — AGENTS.md:27
• [O-002] Eval stimulus name doubles as a natural key for result history (Optional) — cli/evals/azd-app-onboard/eval.yaml:83

[github-actions]

🚀 Test This PR

A preview build (0.20.0-pr609) is ready for testing!

🌐 Website Preview

Live Preview: https://jongio.github.io/azd-app/pr/609/

One-Line Install (Recommended)

PowerShell (Windows):

iex "& { $(irm https://raw.githubusercontent.com/jongio/azd-app/main/cli/scripts/install-pr.ps1) } -PrNumber 609 -Version 0.20.0-pr609"

Bash (macOS/Linux):

curl -fsSL https://raw.githubusercontent.com/jongio/azd-app/main/cli/scripts/install-pr.sh | bash -s 609 0.20.0-pr609

Uninstall

When you're done testing:

PowerShell (Windows):

iex "& { $(irm https://raw.githubusercontent.com/jongio/azd-app/main/cli/scripts/uninstall-pr.ps1) } -PrNumber 609"

Bash (macOS/Linux):

curl -fsSL https://raw.githubusercontent.com/jongio/azd-app/main/cli/scripts/uninstall-pr.sh | bash -s 609

---

Build Info:

• Version: 0.20.0-pr609
• Commit: e112f22
• Release: View pre-release

What to Test:
Please review the PR description and test the changes described there.

[wbreza]

Both generated-artifact defects are fixed at the source, and the cumulative diff is clean.

What stood out

• Fixed at the right layer — The punctuation now lives in proto/azdapp/v1/common.proto:86, and both generated outputs carry the identical semicolon. The Go/TS split is structurally impossible now that one source feeds both.
• Regeneration is genuinely a no-op — The embedded descriptors in common.pb.go and common_pb.ts were generated without source info, so they encode no comment text. Comment-only edits cannot make them stale, which confirms the textual verification was sufficient even without running codegen.
• Doubled colons resolved broadly — Rather than patching the one flagged line, the shape was swept across AGENTS.md, README.md, azure.yaml.md, and the container-orchestration spec, each rewritten to fit the sentence. The two remaining instances in AGENTS.md are unmodified on main and correctly out of scope.
• Wire contract untouched — The edited comment documents append-only enum semantics; the HealthState values themselves are unchanged, and no sanitizer, regex, exec argument, or test assertion moved anywhere in the sweep.

Improvement opportunities

• [O-002] Eval stimulus names double as history keys until an explicit id field exists (Optional) — cli/evals/azd-app-onboard/eval.yaml:83

[wbreza]

Test-only commit that closes the coverage gap honestly, by exercising the branches rather than adjusting the gate.

What stood out

• The dry-run test is not vacuous — The risk with a test like this is that it passes because detection silently found nothing, which would keep it green even if dry-run were broken. It does not: zero detected services returns an error, and the test requires Execute() to succeed. Captured output confirms the Vite fixture detects a service and reaches the dry-run branch.
• Byte-equality is the right assertion here — On the unchanged-configuration branch the file is never opened for write, so exact equality tests real behavior rather than YAML round-trip fidelity, which would have been a brittle claim.
• Both swept strings genuinely execute — Dry run, no files modified and No changes needed, azure.yaml already has complete service configuration both appear in test output, so the two lines the sweep touched are actually covered.
• No process-state hazard — Neither new test calls t.Parallel(), t.Chdir matches the established pattern in this package, NewInitCommand() returns a fresh command with no package-level state, and the full package passes under -race.

Improvement opportunities

• [O-003] Covered lines are executed but their output is never asserted (Optional) — cli/src/cmd/app/commands/init_test.go:348

[wbreza]

[O-003] Both tests execute the two swept cliout.Info(...) lines, confirmed in captured output, but neither asserts what was printed.

The practical consequence: if a future edit reverted Dry run, no files modified to an em-dash form, or otherwise corrupted either string, coverage would stay green and both tests would still pass. So these raise the patch percentage without protecting the exact prose this PR set out to normalize.

That is a legitimate choice if the goal was purely the coverage gate, and the branches themselves are now genuinely exercised, which is the harder half. If you want the strings pinned too, capturing the output stream and asserting the exact ASCII text would close it.

One note if this test ever grows: with t.TempDir() plus t.Chdir and no name in the fixture, the service name is derived from the temp directory basename, so it came through as 001. Harmless for NoFileExists, but it would make any future assertion on generated YAML content flaky. A named subdirectory or an explicit "name" field would make it deterministic.

[wbreza]

The comma splices the sweep introduced are now cleaned up, and the three re-edited user-facing strings have no consumer that breaks.

What stood out

• Right fix for the right problem — Replacing an em dash with a comma between two independent clauses produced splices; converting those to semicolons and colons resolves them without touching meaning. All 22 files read correctly.
• Re-edited strings are safe — Dry run: no files modified, No changes needed: azure.yaml..., and the install_dependencies error each appear exactly once in the tree, at their own definition. No test, eval rubric, must_contain pattern, snapshot, or test-name filter references any of them, so the second edit to these strings breaks nothing.
• Security rationale intact — The CWE-352/CWE-306 comments in server_routes.go, the XSS note in log-utils.ts, and the containment test comments are comment-only and none were weakened. The numbered two-factor list actually reads better with colons.
• Merge blocker is not from this PR — The failing vulnerability check reports only Go standard library advisories in crypto/tls, net/http, net/url, html/template, encoding/asn1, and encoding/xml, all fixed in go1.26.6 against a repo pinned to go1.26.5. The two flagged files this branch touches are comment-only here, so the fix is a toolchain bump, independent of this work.

Improvement opportunities

• [O-004] One comma splice from the original sweep is still unconverted (Optional) — cli/src/internal/service/types.go:466

[wbreza]

[O-004] This line still has the shape the rest of the commit fixed. The original sweep turned an em dash into a comma here:

// runs locally as a process when an explicit local `command` or `type: process`
// is configured, its `docker.*`/image is then used only by `azd deploy`.

Both halves are independent clauses. The second has its own subject and verb (its docker.*/image / is then used), and the then signals exactly the consequence relationship a semicolon carries. It reads as the same splice that mode.go, httputil.go, and checker_process.go all had before this commit.

Comment-only, so there is no behavior impact, and this does not hold anything up. A semicolon in place of the comma would finish the set.