feat: v0.5 supervised internship workbench preview

[vicky-dot20]

Summary

Build the v0.5.0 public-preview scope as a GitHub-native, supervised internship framework.

• One starting page and fork-first onboarding, mentor setup and fork recovery.
• 18 concrete active tasks, isolated submissions and a four-week starter route.
• Six runnable exercise contracts, deliberately incomplete starters, review/debug fixtures and localhost HTTP runner.
• Metadata/dependency/link/catalogue validation, task preparation/checking CLI and regression tests.
• Linux/Windows × Node 22/24 CI, preserving the required validate check and read-only permissions.
• Preserve all 206 IDs; explicitly mark 188 unfinished outlines draft. Correct the initial premature 1.0.0 claim without rewriting history.

Verification

On Windows, Node 22.21.0:

• npm ci: passed.
• npm run validate: passed (206 IDs; 18 active).
• npm test: 37 passed, 0 failed, 0 skipped.
• All six correct reference solutions pass, all six unfinished starters fail.
• CLI integration prepares all 18 tasks, rejects empty evidence, preserves existing work and catches broken submissions.
• npm run check:submissions: passed, zero v0.5 learner submissions committed.
• git diff --cached --check: passed.

GitHub matrix results remain required before merge.

Limitations / review focus

This is not a completed 206-task full-stack curriculum. Advanced tracks and full-stack implementation remain draft. Browser/accessibility outcomes and learner understanding require human review. Prior two-account simulations were AI-assisted, not independent learner validation.

Review active task requirements, training-repo ownership guidance, validation boundaries and Actions permissions. Shared exercise tests are public, not anti-cheating infrastructure.

No Issue is auto-closed; this implements the owner's v0.5 build request.

AI disclosure

AI assistance: Yes. Codex authored this framework implementation, documentation and regression fixtures and ran the commands above. These are maintainer test results, not claims of an intern's independent understanding. Human maintainer review remains pending.

[vkverse]

Reviewed the v0.5 preview scope and changes. All CI checks passed. The distinction between 18 active tasks and 188 draft outlines is clear. Approved for a supervised pilot; independent intern testing remains pending.