docs: refresh CLI workshop with learner-created skills and QA

[GeekTrainer]

Summary

Final layer 4 of 4 in the draft replacement stack for #167, based on #170 (geektrainer/vscode-and-cloud-pr at 7a1715c1e5413ffe1b42ec3194c7eb2f3390ac1c), not main. Mechanically transfers the complete English and five localized CLI trees from pinned revision e150fd531805f57a77be754c37f342386471bc00: three PR milestones, learner-created quality-checks skills, Playwright MCP, requirements-first QA, and bounded CLI controls.

Registered review order: #169 → #174 → #170 → #171. Includes the CLI sidebar, CLI landing descriptions/dates, Spanish and Portuguese locale-local VS Code/Cloud overview links, translated-tree descriptions, and numbered CLI examples. App content is inherited unchanged from #174; VS Code/Cloud lessons and other guidance are untouched by this layer. Current commit 1dcc86f468e883d32de415a87dcf3167f98e4d7d has cumulative tree ffe15594ac09dde2077950e541fb97ad1382dff9. All curriculum bytes remain identical to the preserved complete revision. The full tree differs only in four inherited files from #169 (7407858a6bf8c994073b4876d7e4d99c3dd540e5): the verification skill, skill index, and small Bash/PowerShell wrappers around existing checks. #167 remains open and draft for historical reference.

Type of change

• Workshop content (lesson Markdown, images)
• Site shell (website/ Astro + Starlight wrapper)
• Copilot configuration (.github/copilot-instructions.md, instructions, agents, skills)
• Repo housekeeping (CI, dependabot, README, license)
• Other:

Verification

• Type checks and a clean site build pass.
• Affected pages and translations match their source and intended routes.
• Offline internal link checks pass.
• Changed external links have been checked separately.

Ran the actual bash .github/skills/build-and-verify-docs/scripts/verify.sh once against the final full curriculum at 1dcc86f468e883d32de415a87dcf3167f98e4d7d. It completed npm run check:all, a fresh npm run build, and isolated offline Lychee successfully; wrapper exit 0. Astro reported zero errors/warnings and four existing hints. Offline Lychee reported 0 errors, 2,958 unique links. Reused existing dependencies through temporary package-level links with local caches, without installation or shared dependency mutation; removed the links and local dependency directory afterward. The wrapper cleaned its temporary link root.

These simple wrappers run existing checks; they do not compare source prose, translations, or exact code against rendered HTML. Source/output inspection remains a separate AI review task. Prior inspection of the unchanged curriculum covered all six CLI locales, intended routes, titles/headings, language, prompts, sidebar links, and raw admonition markers. The previously checked seven added external destinations are unchanged. The PowerShell wrapper has not been executed or parser-validated because PowerShell was unavailable.

The 118-file CLI-owned patch is unchanged: the same changed paths and added/deleted patch lines, all 73 CLI file blobs (72 Markdown pages plus the existing .gitkeep), and every curriculum source fence. Verified the exact 184-word English creation prompt and frozen CREATE/RUN SHA-256 hashes. The entire docs/ and website/ trees match the preserved complete snapshot byte-for-byte, including both final sidebars, the original loader, and package manifests. No custom verification engine, integration hooks, or direct parser dependency remains. Only the four inherited skill/index/wrapper files differ from that snapshot. No stale removed CLI-route references or intermediate link blockers were introduced.

Notes for reviewers

Historical evidence, not new application tests: #167 retains the detailed Bash/macOS skill-generation/execution trials with claude-sonnet-5 and mai-code-1.1-flash: lint, unit 35 tests / 3 files, full unfiltered E2E 28 tests / 4 specs, tsgo + Astro type checks, and the documented src/lib/games.test.ts selection 12 tests / 1 file. These are not QA-agent or UI certification. Sonnet recovered read-only permission denials; output-tail pipelines mean an outer zero alone does not prove native exit success. MAI's outer constant-string eval loop was generically denied and never executed, then recovered through five direct Bash calls with EXIT 0; eval was not in the generated wrappers.

Native PowerShell, the broader QA matrix, App agent-picker/hot-discovery behavior, and UI rehearsals remain unverified. No model trials, preview servers, or real-app checks ran for this content partition, reorder, or verification simplification. Existing rendering turns callout bodies into plain divs without raw markers; this partition does not alter shared callout styling.

Template dependencies: github-samples/tailspin-toys#97 and github-samples/tailspin-toys#91. Neither is changed or merged by this PR. Historical #172 was indirectly marked merged into this CLI branch during the reorder; replacement #174 preserves its App review. Nothing merged into main. Keep this PR draft; no auto-merge is enabled.